Sensors

22 pages, 2124 KiB

Open AccessArticle

Online Inductance Monitoring Based on Dynamic Characteristics and ESR Effect Compensation for Buck Converter Without Current Sensor

by Chen Chen, Liang Wang, Wanyang Wang, Run Min and Qiaoling Tong

Sensors 2025, 25(12), 3589; https://doi.org/10.3390/s25123589 - 6 Jun 2025

Abstract

Inductor parameter variations often affect the control performance of digital current mode (CM)-controlled buck converters as their high performance relies on accurate converter modeling. However, recent studies have shown that reliably monitoring inductance with current sensors and high-frequency sampling greatly increases the overall [...] Read more.

Inductor parameter variations often affect the control performance of digital current mode (CM)-controlled buck converters as their high performance relies on accurate converter modeling. However, recent studies have shown that reliably monitoring inductance with current sensors and high-frequency sampling greatly increases the overall cost of this process. To address this issue, an online inductance monitoring method without a current sensor is proposed in this study. First, an inductance calculation model is derived by applying the dynamic characteristics of a buck converter with inductor volt-second and capacitor charge balance principles. The model’s accuracy is guaranteed by considering inductor current switching ripple characteristics. Nevertheless, output capacitor equivalent series resistance (ESR) can degrade the accuracy of the proposed calculation model. Thus, to enhance the tolerance of the inductance calculation model to capacitor ESR, the ESR effect on inductance monitoring is investigated. With the proposed capacitor ESR estimation method, inductance monitoring achieves reliable accuracy, even for a buck converter with high capacitor ESR. The effectiveness of the proposed method is verified by simulations and experiments on a buck converter with digital sensorless current mode (SCM) control. Full article

(This article belongs to the Special Issue Sensors Technologies for Measurements and Signal Processing)

26 pages, 5630 KiB

Open AccessReview

Scientific Machine Learning for Elastic and Acoustic Wave Propagation: Neural Operator and Physics-Guided Neural Network

by Nafisa Mehtaj and Sourav Banerjee

Sensors 2025, 25(12), 3588; https://doi.org/10.3390/s25123588 - 6 Jun 2025

Abstract

Scientific machine learning (SciML) offers an emerging alternative to the traditional modeling approaches for wave propagation. These physics-based models rely on computationally demanding numerical techniques. However, SciML extends artificial neural network-based wave models with the capability of learning wave physics. Contrary to the [...] Read more.

Scientific machine learning (SciML) offers an emerging alternative to the traditional modeling approaches for wave propagation. These physics-based models rely on computationally demanding numerical techniques. However, SciML extends artificial neural network-based wave models with the capability of learning wave physics. Contrary to the physics-intensive methods, particularly physics-informed neural networks (PINNs) presented earlier, this study presents data-driven frameworks of physics-guided neural networks (PgNNs) and neural operators (NOs). Unlike PINNs and PgNNs, which focus on specific PDEs with respective boundary conditions, NOs solve a family of PDEs and hold the potential to easily solve different boundary conditions. Hence, NOs provide a more generalized SciML approach. NOs extend neural networks to map between functions rather than vectors, enhancing their applicability. This review highlights the potential of NOs in wave propagation modeling, aiming to advance wave-based structural health monitoring (SHM). Through comparative analysis of existing NO algorithms applied across different engineering fields, this study demonstrates how NOs improve generalization, accelerate inference, and enhance scalability for practical wave modeling scenarios. Lastly, this article identifies current limitations and suggests promising directions for future research on NO-based methods within computational wave mechanics. Full article

(This article belongs to the Special Issue Feature Review Papers in Physical Sensors)

► Show Figures

Figure 1

24 pages, 5027 KiB

Open AccessReview

Recent Advances in Aptamer-Based Point-of-Care Testing

by Senlin Luo, Xiaotian Zhang, Yuzhuo Zhang, Xiaofeng Qu, Qiru Sun, Tianhuan Peng and Quan Yuan

Sensors 2025, 25(12), 3587; https://doi.org/10.3390/s25123587 - 6 Jun 2025

Abstract

Aptamers refer to a class of oligonucleotide probes that have demonstrated remarkable capabilities beyond mere genetic coding, including the unique ability to recognize and selectively bind to specific molecular targets. Numerous advantages, including accessibility for targeting a diverse array of molecules and compatibility [...] Read more.

Aptamers refer to a class of oligonucleotide probes that have demonstrated remarkable capabilities beyond mere genetic coding, including the unique ability to recognize and selectively bind to specific molecular targets. Numerous advantages, including accessibility for targeting a diverse array of molecules and compatibility with different signal amplification and transduction elements, underscore the application of aptamers for delivering rapid and accurate diagnostic tests at the point of care. This review provides a comprehensive summary of the recent advances in aptamer-based point-of-care testing, especially highlighting the innovative applications of aptamers in colorimetric sensors, lateral flow assays, fluorescent biosensors, and electrochemical biosensors. Additionally, current challenges in this burgeoning field and forward-looking perspectives for aptamer-based point-of-care testing are discussed. Full article

(This article belongs to the Special Issue Biosensors for Point-of-Care Diagnostics)

34 pages, 32280 KiB

Open AccessArticle

Closing the Domain Gap: Can Pseudo-Labels from Synthetic UAV Data Enable Real-World Flood Segmentation?

by Georgios Simantiris, Konstantinos Bacharidis and Costas Panagiotakis

Sensors 2025, 25(12), 3586; https://doi.org/10.3390/s25123586 - 6 Jun 2025

Abstract

We present a novel methodology for generating and filtering synthetic Unmanned Aerial Vehicle (UAV) flood imagery to enhance the generalization capabilities of segmentation models. Our framework combines text-to-image synthesis and image inpainting, using curated prompts and real-world segmentation masks to produce diverse and [...] Read more.

We present a novel methodology for generating and filtering synthetic Unmanned Aerial Vehicle (UAV) flood imagery to enhance the generalization capabilities of segmentation models. Our framework combines text-to-image synthesis and image inpainting, using curated prompts and real-world segmentation masks to produce diverse and realistic flood scenes. To overcome the lack of human annotations, we employ an unsupervised pseudo-labeling method that generates segmentation masks based on floodwater appearance characteristics. We further introduce a filtering stage based on outlier detection in feature space to improve the realism of the synthetic dataset. Experimental results on five state-of-the-art flood segmentation models show that synthetic data can closely match real data in training performance, and combining both sources improves model robustness by 1–7%. Finally, we investigate the impact of prompt design on the visual fidelity of generated images and provide qualitative and quantitative evidence of distributional similarity between real and synthetic data. Full article

(This article belongs to the Special Issue Sensors and Advanced Sensing Techniques for Computer Vision Applications: Second Edition)

► Show Figures

Graphical abstract

25 pages, 2350 KiB

Open AccessArticle

A Robust Rotation-Equivariant Feature Extraction Framework for Ground Texture-Based Visual Localization

by Yuezhen Cai, Linyuan Xia, Ting On Chan, Junxia Li and Qianxia Li

Sensors 2025, 25(12), 3585; https://doi.org/10.3390/s25123585 - 6 Jun 2025

Abstract

Ground texture-based localization leverages environment-invariant, planar-constrained features to enhance pose estimation robustness, thus offering inherent advantages for seamless localization. However, traditional feature extraction methods struggle with reliable performance under large-scale rotations and texture sparsity in the case of ground texture-based localization. This study [...] Read more.

Ground texture-based localization leverages environment-invariant, planar-constrained features to enhance pose estimation robustness, thus offering inherent advantages for seamless localization. However, traditional feature extraction methods struggle with reliable performance under large-scale rotations and texture sparsity in the case of ground texture-based localization. This study addresses these challenges through a learning-based feature extraction framework—Ground Texture Rotation-Equivariant Keypoints and Descriptors (GT-REKD). The GT-REKD framework employs group-equivariant convolutions over the cyclic rotation group, augmented with directional attention and orientation-encoding heads, to produce dense keypoints and descriptors that are exactly invariant to 0–360° in-plane rotations. The experimental results for ground texture localization show that GT-REKD achieves 96.14% matching in pure rotation tests, 94.08% in incremental localization, and relocalization errors of 5.55° and 4.41 px (≈0.1 cm), consistently outperforming baseline methods under extreme rotations and sparse textures, highlighting its applicability to visual localization and simultaneous localization and mapping (SLAM) tasks. Full article

(This article belongs to the Section Navigation and Positioning)

18 pages, 6606 KiB

Open AccessArticle

Method for Estimating Amount of Saliva Secreted Using a Throat Microphone

by Kai Washino, Ayumi Ohnishi, Tsutomu Terada and Masahiko Tsukamoto

Sensors 2025, 25(12), 3584; https://doi.org/10.3390/s25123584 - 6 Jun 2025

Abstract

Saliva is an important secretion, and a continued insufficient amount of saliva secreted causes glossitis, stomatitis, and so on. Since the amount of saliva secreted changes daily, adverse effects occur daily. Therefore, it is necessary to constantly measure the amount of saliva secreted [...] Read more.

Saliva is an important secretion, and a continued insufficient amount of saliva secreted causes glossitis, stomatitis, and so on. Since the amount of saliva secreted changes daily, adverse effects occur daily. Therefore, it is necessary to constantly measure the amount of saliva secreted and take appropriate measures when it decreases. However, there is no method to constantly measure saliva. We propose a method to estimate the amount of saliva secreted from the sound acquired by a wearable throat microphone. The proposed method uses deep learning to classify whether the sound acquired by the throat microphone is swallowing or not. Based on the swallowing information, the proposed method estimates the amount of saliva secreted. The accuracy of the classification of swallowing was 96.96%. For the estimation of the amount of saliva secreted, the R was 0.600 and

M A E

was

0.0487

. Full article

(This article belongs to the Special Issue Wearable Sensing of Medical Condition at Home Environment)

34 pages, 2032 KiB

Open AccessReview

The IoT and AI in Agriculture: The Time Is Now—A Systematic Review of Smart Sensing Technologies

by Tymoteusz Miller, Grzegorz Mikiciuk, Irmina Durlik, Małgorzata Mikiciuk, Adrianna Łobodzińska and Marek Śnieg

Sensors 2025, 25(12), 3583; https://doi.org/10.3390/s25123583 - 6 Jun 2025

Abstract

The integration of the Internet of Things (IoT) and artificial intelligence (AI) has reshaped modern agriculture by enabling precision farming, real-time monitoring, and data-driven decision-making. This systematic review, conducted in accordance with the PRISMA methodology, provides a comprehensive overview of recent advancements in [...] Read more.

The integration of the Internet of Things (IoT) and artificial intelligence (AI) has reshaped modern agriculture by enabling precision farming, real-time monitoring, and data-driven decision-making. This systematic review, conducted in accordance with the PRISMA methodology, provides a comprehensive overview of recent advancements in smart sensing technologies for arable crops and grasslands. We analyzed the peer-reviewed literature published between 2020 and 2024, focusing on the adoption of IoT-based sensor networks and AI-driven analytics across various agricultural applications. The findings reveal a significant increase in research output, particularly in the use of optical, acoustic, electromagnetic, and soil sensors, alongside machine learning models such as SVMs, CNNs, and random forests for optimizing irrigation, fertilization, and pest management strategies. However, this review also identifies critical challenges, including high infrastructure costs, limited interoperability, connectivity constraints in rural areas, and ethical concerns regarding transparency and data privacy. To address these barriers, recent innovations have emphasized the potential of Edge AI for local inference, blockchain systems for decentralized data governance, and autonomous platforms for field-level automation. Moreover, policy interventions are needed to ensure fair data ownership, cybersecurity, and equitable access to smart farming tools, especially in developing regions. This review is the first to systematically examine AI-integrated sensing technologies with an exclusive focus on arable crops and grasslands, offering an in-depth synthesis of both technological progress and real-world implementation gaps. Full article

(This article belongs to the Special Issue Smart Sensing Systems for Arable Crop and Grassland Management)

25 pages, 38512 KiB

Open AccessArticle

A Novel Audio-Perception-Based Algorithm for Physiological Monitoring

by Zixuan Zhang, Wenxuan Jin, Dejiao Huang and Zhongwei Sun

Sensors 2025, 25(12), 3582; https://doi.org/10.3390/s25123582 - 6 Jun 2025

Abstract

Exercise metrics are critical for assessing health, but real-time heart rate and respiration measurements remain challenging. We propose a physiological monitoring system that uses an in-ear microphone to extract heart rate and respiration from faint ear canal signals. An improved non-negative matrix factorization [...] Read more.

Exercise metrics are critical for assessing health, but real-time heart rate and respiration measurements remain challenging. We propose a physiological monitoring system that uses an in-ear microphone to extract heart rate and respiration from faint ear canal signals. An improved non-negative matrix factorization (NMF) algorithm combines with a short-time Fourier transform (STFT) to separate physiological components, while an inverse Fourier transform (IFT) reconstructs the signal. The earplug effect enhances the low-frequency components, thereby improving the signal quality and noise immunity. Heart rate is derived from short-term energy and zero-crossing rate, while a BiLSTM-based model can refine the breathing phases and calculate indicators such as respiratory rate. Experiments have shown that the average accuracy can reach 91% under various conditions, exceeding 90% in different environments and under different weights, thus ensuring the system’s robustness. Full article

(This article belongs to the Section Physical Sensors)

11 pages, 3189 KiB

Open AccessArticle

Robotic-Assisted XRF Testing System for In-Situ Areal Density Measurement of Light-Sensitive Explosive Coatings

by Chang Xu, Haibin Xu, Ke Wu, Bo Chen, Pengju Dong, Yaguang Sui and Hai Chen

Sensors 2025, 25(12), 3581; https://doi.org/10.3390/s25123581 - 6 Jun 2025

Abstract

The light-sensitive explosive (silver acetylide–silver nitrate, SASN) sprayed on structural surfaces can be synchronously initiated by intense pulsed flash, thereby simulating cold X-ray blow-off events characterized by thermal–mechanical coupling effects. By adjusting the areal density of SASN coatings, proportional blow-off impulse levels can [...] Read more.

The light-sensitive explosive (silver acetylide–silver nitrate, SASN) sprayed on structural surfaces can be synchronously initiated by intense pulsed flash, thereby simulating cold X-ray blow-off events characterized by thermal–mechanical coupling effects. By adjusting the areal density of SASN coatings, proportional blow-off impulse levels can be achieved. To address the challenge of in situ and non-destructive areal density measurement for SASN coatings, this study developed an X-ray fluorescence (XRF) detection system integrated with a six-axis spray robot. Excitation parameters (50 kV, 20 μA) and geometric configuration (6 cm focal distance) were optimized to establish a quadratic calibration model between Ag K_α counts and areal density (0–80 mg/cm²) with high correlation (R² = 0.9987). Validation experiments were conducted on a uniformly coated SASN plate (20 × 20 cm) to evaluate the consistency between XRF and sampling methods. The XRF-measured areal density averaged 12.722 mg/cm² with a coefficient of variation (CV) of 3.19%. The reference value obtained by the sampling method was 12.718 mg/cm² (CV = 1.57%). The relative deviation between the two methods was only 0.03%, confirming the feasibility of XRF for the quantification of SASN coatings. The XRF system completed measurements in 1 h, achieving a 77.8% time reduction compared to conventional sampling (4.5 h), significantly enhancing efficiency. This work provides a reliable solution for in situ and non-destructive quality control of energetic material coatings. Full article

(This article belongs to the Section Physical Sensors)

► Show Figures

Figure 1

17 pages, 3374 KiB

Open AccessArticle

Gramian Angular Field and Convolutional Neural Networks for Real-Time Multiband Spectrum Sensing in Cognitive Radio Networks

by Yanqueleth Molina-Tenorio, Alfonso Prieto-Guerrero, Enrique Rodriguez-Colina, Luis Alberto Vásquez-Toledo and Omar Alejandro Olvera-Guerrero

Sensors 2025, 25(12), 3580; https://doi.org/10.3390/s25123580 - 6 Jun 2025

Abstract

Multiband spectrum sensing in a cooperative environment is a novel solution for efficient spectrum resource management under the cognitive radio networks (CRNs) paradigm. This paper presents a distinctive framework where a central entity collects power spectral density data from multiple geographically distributed secondary [...] Read more.

Multiband spectrum sensing in a cooperative environment is a novel solution for efficient spectrum resource management under the cognitive radio networks (CRNs) paradigm. This paper presents a distinctive framework where a central entity collects power spectral density data from multiple geographically distributed secondary users and applies the Gramian angular field (GAF) summation method to transform the time-series data into image representations. A major contribution of this work is the integration of these GAF images with a convolutional neural network (CNN), enabling precise and real-time detection of primary user activity and spectrum occupancy. The proposed approach achieves 99.6% accuracy in determining spectrum occupancy, significantly outperforming traditional sensing techniques. The main contributions of this study are (i) the introduction of GAF-based image representations for cooperative spectrum sensing in CRNs; (ii) the development of a CNN-based classification framework for enhanced spectrum occupancy detection; and (iii) the demonstration of superior detection performance in dynamic, real-time environments. Full article

(This article belongs to the Special Issue Advances in Remote Sensing and Electromagnetic Spectrum Sensing: Data Acquisition and Signal Processing)

► Show Figures

Figure 1

21 pages, 7778 KiB

Open AccessArticle

High-Resolution Localization Using Distributed MIMO FMCW Radars

by Huijea Park, Seungsu Chung, Jaehyun Park and Yang Huang

Sensors 2025, 25(12), 3579; https://doi.org/10.3390/s25123579 - 6 Jun 2025

Abstract

Due to its fast processing time and robustness against harsh environmental conditions, the frequency modulated continuous waveform (FMCW) multiple-input multiple-output (MIMO) radar is widely used for target localization. For high-accuracy localization, the two-dimensional multiple signal classification (2D MUSIC) algorithm can be applied to [...] Read more.

Due to its fast processing time and robustness against harsh environmental conditions, the frequency modulated continuous waveform (FMCW) multiple-input multiple-output (MIMO) radar is widely used for target localization. For high-accuracy localization, the two-dimensional multiple signal classification (2D MUSIC) algorithm can be applied to signals received by a single FMCW MIMO radar, achieving high-resolution positioning performance. To further enhance estimation accuracy, received signals or MUSIC spectra from multiple FMCW MIMO radars are often collected at a data fusion center and processed coherently. However, this approach increases data communication overhead and implementation complexity. To address these challenges, we propose an efficient high-resolution target localization algorithm. In the proposed method, the target position estimates from multiple FMCW MIMO radars are collected and combined using a weighted averaging approach to determine the target’s position within a unified coordinate system at the data fusion center. We first analyze the achievable resolution in the unified coordinate system, considering the impact of local parameter estimation errors. Based on this analysis, weights are assigned according to the achievable resolution within the unified coordinate framework. Notably, due to the typically limited number of antennas in FMCW MIMO radars, the azimuth angle resolution tends to be relatively lower than the range resolution. As a result, the achievable resolution in the unified coordinate system depends on the placement of each FMCW MIMO radar. The performance of the proposed scheme is validated using both synthetic simulation data and experimentally measured data, demonstrating its effectiveness in real-world scenarios. Full article

(This article belongs to the Special Issue Feature Papers in the 'Sensor Networks' Section 2025)

► Show Figures

Figure 1

14 pages, 3586 KiB

Open AccessArticle

Design and Analysis of Differential Compensated Eddy Current Displacement Sensors

by Yuliang Bian, Kun Zhang and Tiehua Ma

Sensors 2025, 25(12), 3578; https://doi.org/10.3390/s25123578 - 6 Jun 2025

Abstract

The traditional eddy current displacement sensor is limited by the diameter of the probe coil, and the measuring range is relatively small. In order to improve the range of an eddy current displacement sensor, a differential compensated eddy current displacement sensor (DCECDS) is [...] Read more.

The traditional eddy current displacement sensor is limited by the diameter of the probe coil, and the measuring range is relatively small. In order to improve the range of an eddy current displacement sensor, a differential compensated eddy current displacement sensor (DCECDS) is proposed in this paper. The DCECDS is designed with three coils and is characterized by a large measuring range and good linearity. Based on the analysis of the working principle of DCECDS, the corresponding equivalent circuit model and mathematical model were established in this study. The effects of excitation frequency, inner diameter and thickness of excitation coil on the sensor performance were investigated. Based on the simulation results, the structural parameters of the sensor were designed and verified by experiments. The experimental results show that the measuring range of the DCECDS designed in this study can reach 1.76 times the diameter of the probe coil, which is 3.5 times higher than that of the traditional eddy current displacement sensor. The measuring range of the eddy current displacement sensor is effectively improved, which provides an important reference and practical value for the design of eddy current displacement sensors. Full article

(This article belongs to the Section Physical Sensors)

► Show Figures

Figure 1

20 pages, 3449 KiB

Open AccessReview

Bayesian Network in Structural Health Monitoring: Theoretical Background and Applications Review

by Qi-Ang Wang, Ao-Wen Lu, Yi-Qing Ni, Jun-Fang Wang and Zhan-Guo Ma

Sensors 2025, 25(12), 3577; https://doi.org/10.3390/s25123577 - 6 Jun 2025

Abstract

With accelerated urbanization and aging infrastructure, the safety and durability of civil engineering structures face significant challenges, making structural health monitoring (SHM) a critical approach to ensuring engineering safety. The Bayesian network, as a probabilistic reasoning tool, offers a novel technological pathway for [...] Read more.

With accelerated urbanization and aging infrastructure, the safety and durability of civil engineering structures face significant challenges, making structural health monitoring (SHM) a critical approach to ensuring engineering safety. The Bayesian network, as a probabilistic reasoning tool, offers a novel technological pathway for SHM due to its strengths in handling uncertainties and multi-source data fusion. This study systematically reviews the core applications of the Bayesian network in SHM, including damage prediction, data fusion, uncertainty modeling, and decision support. By integrating multi-source sensor data with probabilistic inference, the Bayesian network enhances the accuracy and reliability of monitoring systems, providing a theoretical foundation for damage identification, risk early warning, and optimization of maintenance strategies. The study presents a comprehensive review that systematically unifies the theoretical framework of BN with SHM applications, addressing the gap between probabilistic reasoning and real-world infrastructure management. The research outcomes hold significant theoretical and engineering implications for advancing SHM technology development, reducing operational and maintenance costs, and ensuring the safety of public infrastructure. Full article

(This article belongs to the Special Issue Condition Assessment and Reliability Analysis Enabled by Structural Health Monitoring Sensors)

► Show Figures

Figure 1

19 pages, 2465 KiB

Open AccessArticle

The Design and Implementation of a Dynamic Measurement System for a Large Gear Rotation Angle Based on an Extended Visual Field

by Po Du, Zhenyun Duan, Jing Zhang, Wenhui Zhao, Engang Lai and Guozhen Jiang

Sensors 2025, 25(12), 3576; https://doi.org/10.3390/s25123576 - 6 Jun 2025

Abstract

High-precision measurement of large gear rotation angles is a critical technology in gear meshing-based measurement systems. To address the challenge of high-precision rotation angle measurement for large gear, this paper proposes a binocular vision method. The methodology consists of the following steps: First, [...] Read more.

High-precision measurement of large gear rotation angles is a critical technology in gear meshing-based measurement systems. To address the challenge of high-precision rotation angle measurement for large gear, this paper proposes a binocular vision method. The methodology consists of the following steps: First, sub-pixel edges of calibration circles on a 2D dot-matrix calibration board are extracted using edge detection algorithms to obtain pixel coordinates of the circle centers. Second, a high-precision calibration of the measurement reference plate is achieved through a 2D four-parameter coordinate transformation algorithm. Third, binocular cameras capture images of the measurement reference plates attached to large gear before and after rotation. Coordinates of the camera’s field-of-view center in the measurement reference plate coordinate system are calculated via image processing and rotation angle algorithms, thereby determining the rotation angle of the large gear. Finally, a binocular vision rotation angle measurement system was developed, and experiments were conducted on a 600 mm-diameter gear to validate the feasibility of the proposed method. The results demonstrate a measurement accuracy of 7 arcseconds (7”) and a repeatability precision of 3 arcseconds (3”) within the 0–30° rotation range, indicating high accuracy and stability. The proposed method and system effectively meet the requirements for high-precision rotation angle measurement of large gear. Full article

(This article belongs to the Section Physical Sensors)

► Show Figures

Figure 1

20 pages, 10186 KiB

Open AccessArticle

SC-CoSF: Self-Correcting Collaborative and Co-Training for Image Fusion and Semantic Segmentation

by Dongrui Yang, Lihong Qiao and Yucheng Shu

Sensors 2025, 25(12), 3575; https://doi.org/10.3390/s25123575 - 6 Jun 2025

Abstract

Multimodal image fusion and semantic segmentation play pivotal roles in autonomous driving and robotic systems, yet their inherent interdependence remains underexplored. To address this gap and overcome performance bottlenecks, we propose SC-CoSF, a novel coupled framework that jointly optimizes these tasks through synergistic [...] Read more.

Multimodal image fusion and semantic segmentation play pivotal roles in autonomous driving and robotic systems, yet their inherent interdependence remains underexplored. To address this gap and overcome performance bottlenecks, we propose SC-CoSF, a novel coupled framework that jointly optimizes these tasks through synergistic learning. Our approach replaces traditional duplex encoders with a weight-sharing CNN encoder, implicitly aligning multimodal features while reducing parameter overhead. The core innovation lies in our Self-correction and Collaboration Fusion Module (Sc-CFM), which integrates (1) a Self-correction Long-Range Relationship Branch (Sc-LRB) to strengthen global semantic modeling, (2) a Self-correction Fine-Grained Branch (Sc-FGB) for enhanced visual detail retention through local feature aggregation, and (3) a Dual-branch Collaborative Recalibration (DCR) mechanism for cross-task feature refinement. This design preserves critical edge textures and color contrasts for segmentation while leveraging segmentation-derived spatial priors to guide fusion. We further introduce the Interactive Context Recovery Mamba Decoder (ICRM) to restore lost long-range dependencies during the upsampling process; meanwhile, we propose the Region Adaptive Weighted Reconstruction Decoder (ReAW), which is mainly used to reduce feature redundancy in image fusion tasks. End-to-end joint training enables gradient propagation across all task branches via shared parameters, exploiting inter-task consistency for superior performance. Experiments demonstrate significant improvements over independently optimized baselines in both fusion quality and segmentation accuracy. Full article

(This article belongs to the Section Sensing and Imaging)

► Show Figures

Figure 1

16 pages, 3114 KiB

Open AccessArticle

TDA-L: Reducing Latency and Memory Consumption of Test-Time Adaptation for Real-Time Intelligent Sensing

by Rahim Hossain, Md Tawheedul Islam Bhuian and Kyoung-Don Kang

Sensors 2025, 25(12), 3574; https://doi.org/10.3390/s25123574 - 6 Jun 2025

Abstract

Vision–language models learn visual concepts from the supervision of natural language. It can significantly enhance the generalizability of real-time intelligent sensing, such as analyzing camera-captured real-time images for visually impaired users. However, adapting vision–language models to distribution shifts at test time, caused by [...] Read more.

Vision–language models learn visual concepts from the supervision of natural language. It can significantly enhance the generalizability of real-time intelligent sensing, such as analyzing camera-captured real-time images for visually impaired users. However, adapting vision–language models to distribution shifts at test time, caused by several factors such as lighting or weather changes, remains challenging. In particular, most existing test-time adaptation methods rely on gradient-based fine-tuning and backpropagation, making them computationally expensive and unsuitable for real-time applications. To address this challenge, the Training-Free Dynamic Adapter (TDA) has recently been introduced as a lightweight alternative that uses a dynamic key–value cache and pseudo-label refinement for test-time adaptation without backpropagation. Building on this, we propose TDA-L, a new framework that integrates Low-Rank Adaptation (LoRA) to reduce the size of feature representations and related computational overhead at test time using pre-learned low-rank matrices. TDA-L applies LoRA transformations to both query and cached features during inference, cost-efficiently improving robustness to distribution shifts while maintaining the training-free nature of TDA. Experimental results on seven benchmarks show that TDA-L maintains accuracy but achieves lower latency, less memory consumption, and higher throughput, making it well-suited for AI-based real-time sensing. Full article

(This article belongs to the Special Issue Edge AI for Wearables and IoT)

► Show Figures

Figure 1

15 pages, 5492 KiB

Open AccessReview

Secure and Trusted Crowdsensing for Outdoor Air Quality Monitoring: State of the Art and Perspectives

by Claudio Marche, Emmanuele Massidda, Alessandro Sanna, Gianmarco Angius, Michele Nitti, Davide Maiorca and Stefano Lai

Sensors 2025, 25(12), 3573; https://doi.org/10.3390/s25123573 - 6 Jun 2025

Abstract

Air pollution is a major problem in the modern world; although it particularly impacts developing countries, which are experiencing fast and often uncontrolled industrialization, its effects constitute a global burden on the environment and health. At the same time, the costs of effective [...] Read more.

Air pollution is a major problem in the modern world; although it particularly impacts developing countries, which are experiencing fast and often uncontrolled industrialization, its effects constitute a global burden on the environment and health. At the same time, the costs of effective air quality monitoring programs are prohibitive for emerging economies, thus making any correction difficult to assess. Emerging technologies, such as distributed networks of sensors organized in the Internet of Things, are under the lens of scientific and industrial communities as a valuable, low-cost alternative to standard techniques. In this paper, we report a review of current approaches to distributed air quality monitoring. Specifically, we (1) emphasize the role of crowdsensing in leveraging sensor-enabled mobile devices for large-scale environmental data collection and (2) discuss criticalities, open challenges, and future perspectives in enforcing data security when such approaches are deployed in real application scenarios. Full article

(This article belongs to the Section Physical Sensors)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Sensors, Volume 25, Issue 12 (June-2 2025) – 17 articles

Further Information

Guidelines

MDPI Initiatives

Follow MDPI