This work leverages the continuous sweeping motion of LiDAR scanning to concentrate object detection efforts on specific regions that receive a change in point data from one frame to another. We achieve this by using a sliding time window with short strides and consider the temporal dimension by storing convolution results between passes. This allows us to ignore unchanged regions, significantly reducing the number of convolution operations per forward pass without sacrificing accuracy. This data reuse scheme introduces extreme sparsity to detection data. To exploit this sparsity, we extend our previous work on scatter-based convolutions to allow for data reuse, and as such propose Sparse Scatter-Based Convolution Algorithm with Temporal Data Recycling (SSCATeR). This operation treats incoming LiDAR data as a continuous stream and acts only on the changing parts of the point cloud. By doing so, we achieve the same results with as much as a 6.61-fold reduction in processing time. Our test results show that the feature maps output by our method are identical to those produced by traditional sparse convolution techniques, whilst greatly increasing the computational efficiency of the network.

Hyperspectral image (HSI) classification is central to environmental monitoring, yet real-time deployment of deep learning models on resource-constrained edge platforms remains challenging due to high spectral dimensionality and computational overhead. In this paper, we propose TinyCapsViT, an ultra-lightweight hybrid architecture that integrates convolutional feature extraction, transformer-based attention, and capsule-inspired representation learning for efficient HSI classification under strict TinyML constraints. The model employs a minimal convolutional stem using pointwise and depthwise separable convolutions to capture local spectral-spatial features, followed by a compact tokenization strategy and learnable positional embeddings. A lightweight self-attention module enables global context modeling with reduced computational complexity, while a capsule-inspired refinement block with squash nonlinearity and residual scaling enhances feature discrimination. The proposed architecture contains only 2781 trainable parameters, representing up to a 30× reduction compared with larger architectures such as ResNet and Vision Transformer, and approximately 11× and 9.5× fewer parameters than the CNN and 3D-CNN baselines, respectively. Extensive experiments across benchmark hyperspectral datasets demonstrate that TinyCapsViT achieves up to 99.61% overall accuracy and maintains competitive classification performance despite its substantially reduced model complexity. Although TinyCapsViT does not consistently achieve the highest classification accuracy compared with larger baseline models, it provides a favourable trade-off between classification performance and computational efficiency. These results demonstrate the potential of TinyCapsViT as a practical solution for real-time hyperspectral analysis on resource-constrained edge platforms and UAV-based environmental monitoring applications.

Hyperspectral image (HSI) classification is vital for environmental monitoring, land cover mapping, and precision agriculture, but its effectiveness is often constrained by the scarcity of labeled samples and high spectral similarity among classes. To address these challenges, we propose CrossCapsViT, a hybrid classification framework that integrates capsule networks (CapsNets) and vision transformers (ViTs) through a cross-attention fusion mechanism and a cross-layer adaptive fusion module, enabling richer and more discriminative spectral-spatial feature learning. To further improve efficiency in data scarce scenarios, we embed an actor-critic reinforcement learning-based active learning (RAL) strategy that jointly leverages accuracy, uncertainty, and diversity in the reward structure, guiding the selection of the most informative samples while reducing labeling effort. Experiments conducted on four benchmark datasets (Kennedy Space Center, Pavia University, Houston University 2013, and Salinas) and a custom UAV-based saltmarsh dataset (Derrymore, collected with a Pika-L sensor) demonstrate that CrossCapsViT with RAL consistently outperforms CapsViT and other baseline models in terms of classification accuracy, robustness, and generalizability. The proposed framework achieves up to 25% improvement in class-level accuracy on challenging vegetation classes, while reducing dependence on large annotated datasets, highlighting its potential for practical deployment in real-world ecological monitoring and remote sensing applications.

Rotation invariance is essential for precise object-level segmentation in UAV aerial imagery, where targets can have arbitrary orientations and exhibit fine-scale details. Conventional segmentation architectures like UNet rely on convolution operators that are not rotation-invariant, leading to degraded segmentation accuracy across varying viewpoints. Rotation invariance can be achieved by expanding the filter bank across multiple orientations; however, this significantly increases computational cost and memory requirements. In this article, we introduce a Graphics Processing Unit (GPU)-optimized rotation-invariant convolution framework that eliminates the traditional data lowering (im2col) step required for matrix multiplication-based convolution. By exploiting structured data sharing among symmetrically rotated filters, our method achieves multiorientation convolution with greatly reduced memory requirements and computational redundancy. We further generalize the approach to accelerate convolution with arbitrary (nonsymmetric) rotation angles. Integrated into a UNet segmentation model, the framework yields up to a 5.7% improvement in accuracy over the nonrotation-aware baseline. Across extensive benchmarks, the proposed convolution achieves 20%-57% faster training and 15%-45% lower energy consumption than cuDNN, while maintaining accuracy comparable to state-of-the-art rotation-invariant methods. Because the scatter-based operator greatly reduces intermediate feature dimensionality, the efficiency of our design also enables practical 16-orientation convolution and pooling, yielding further accuracy gains that are infeasible for conventional rotation-invariant implementations. Our 16-orientation approach achieves competitive accuracy on multiple datasets compared with state-of-the-art UAV segmentation networks. These results demonstrate that the proposed method provides an effective and efficient alternative to existing rotation-invariant convolution frameworks.

Unmanned Aerial Vehicle (UAV) remote sensing has gained increasing attention in the scientific community and has rapidly evolved into a widely used tool for diverse applications, particularly in coastal environment monitoring. This study presents a comprehensive review of UAV-based coastal ecosystem monitoring by analysing 1,972 research articles published between 2020 and 2024. Following the PRISMA framework, 406 articles were systematically selected, from which 100 studies underwent detailed technical and ecological analysis. The review critically evaluates UAV platforms, sensor technologies, ecological applications, spatial resolutions, analytical algorithms, field validation approaches, software tools, and observed limitations. The study further provides an in-depth discussion on the current status, emerging trends, and technological advancements in the field, along with recommendations and research directions. Key findings reveal that multirotor platforms with RGB cameras remain dominant, while there is a clear shift towards multispectral, hyperspectral, and LiDAR integration. Additionally, the standardisation of SfM-MVS photogrammetric workflows and the increasing use of RTK/PPK positioning systems are apparent, although GCP-based validation still remains common. The analytical landscape has evolved toward automated machine learning and deep learning frameworks, though weak model interpretability remains a persistent bottleneck. UAVs demonstrated clear advantages for fine-scale ecological mapping, event-driven monitoring, and surveys in inaccessible environments, while geometric accuracy assessment was consistently prioritised in the field validation. Emerging opportunities include sensor/model fusion, explainable AI integration, and new ecological applications such as carbon flux estimation. Hence, this review provides a comprehensive foundation for researchers to effectively integrate UAVs into coastal monitoring applications and identify future research directions.

This paper presents a proof-of-concept investigation into a novel hermetically sealed tunable-medium Extrinsic Fabry-Pérot Interferometer (EFPI) temperature sensor architecture. A series of tuneable-sensitivity EFPI temperature sensors is demonstrated, comprising a large-diameter fused silica diaphragm with a 800 μm diameter, significantly exceeding conventional designs (typically ∼125 μm), with polished diaphragm thicknesses ranging from 28 to 49 μm, housed in hermetically sealed rigid melting point capillaries with a 1.8 mm internal diameter. By exploiting thermally induced pressure differentials generated by a tunable Krytox GPL 105 oil/air fill fraction within the sealed rigid cavity, the sensors demonstrate a continuously tuneable sensitivity design space spanning 0.45 to 190 nm/K. An exact nonlinear thermal pressure model is derived and validated, replacing the linearised approximation which is shown to be inapplicable at fill fractions approaching unity. The low-sensitivity configuration (0.45 nm/K) was characterised at the National Standards Authority of Ireland (NSAI) National Metrology Laboratory against ITS-90 fixed points: the Triple Point of Water (273.16 K) and the Gallium Fixed Point (302.9146 K), with traceability to the International Temperature Scale of 1990 (ITS-90), yielding an instrument-limited resolution of <1.1 mK, consistent with the metrological validation environment. The high-sensitivity configurations (21 and 190 nm/K) were characterised on a laboratory bench, achieving instrument-limited theoretical resolutions of <24 μK and <2.6 μK respectively, pending future metrological validation. The 190 nm/K sensitivity represents an improvement of approximately 21.7× over the closest directly comparable prior Citationutilised fusion splicing and manual polishing. Future development priorities include metrological validation of the high-sensitivity configurations, long-term stability characterisation, thermal cycling, and progression towards an all-glass hermetically sealed construction.

Contaminants of emerging concern (CECs), including pharmaceuticals, pesticides, and PFAS, have attracted increased attention due to their potential to affect the environment and human health. At the same time, environmental DNA (eDNA) can detect and monitor biological communities and can complement chemical monitoring to give a more comprehensive picture of ecosystem status. The simultaneous sampling of CECs and eDNA presents significant technical and logistical challenges and requires very sensitive techniques. Autonomous surface vehicles (ASVs) offer a flexible platform for monitoring coastal water systems, particularly when repeated or prolonged sampling is required. Their use is increasingly relevant for supporting emerging biological and chemical monitoring techniques. Despite its potential, few studies investigate seawater ecosystems using this combined approach. This work involves innovative monitoring of Irish coastal waters using an interdisciplinary approach that integrates expertise in engineering, chemistry, and biology. Research involving an ASV capable of reliable dynamic positioning during extended sampling operations will be shown alongside sensitive analytical techniques for investigating CECs and eDNA in seawater matrices. Results will show strategies to address a key challenge for ASV-based eDNA sampling of maintaining precise station for adequate periods while water is actively pumped through our filtration systems. Study observations include methods for sample handling to overcome the challenge of low target analyte concentration degradation, and contamination.

The conservation of aquatic ecosystems is essential to protect biodiversity, sustain ecological balance, and safeguard human health, particularly as pollution levels continue to rise. Floating plastic debris presents a significant threat to these environments, impacting marine life and degrading water quality. Unmanned Aerial Vehicles (UAVs) have emerged as a scalable and cost-effective alternative for detecting and tracking floating waste across large water bodies. This study presents a UAV framework that performs live detection and geolocation of floating debris in parallel with flight operations. The system integrates DJI Mobile SDK (MSDK) and Robot Operating System (ROS) to enable autonomous navigation, telemetry acquisition, and live video streaming. A YOLOv8 object detection model, trained on a combination of public and custom datasets, performs inference on the video stream to identify debris, while geolocation is calculated from synchronized GPS and IMU data published via ROS. Field trials demonstrated high detection precision at altitudes up to 7.5 meters, with latency inference times under 1.5 seconds per frame and stable performance across sunny and cloudy conditions. The geolocation pipeline achieved consistently accurate positioning, with detections closely matching their actual locations in satellite imagery. The proposed framework is modular and compatible with multiple UAV configurations, allowing integration across different platforms with available telemetry and video streams. It also supports future extensions such as adaptive path planning and coordination with collection systems, including the deployment of autonomous surface vehicles to retrieve detected waste. The results demonstrate the potential of the system as a reliable and adaptable solution for autonomous aquatic waste monitoring.

This paper advances the development of Extrinsic Fabry–Pérot interferometry (EFPI) for high-precision pressure sensing. Presented is an EFPI featuring a diameter of 800 μm with a 7.4 μm diaphragm thickness, demonstrating a resolution of 3.35 mPa and a sensitivity of 149 nm/kPa positioning it amongst the most sensitive fibre optic pressure sensors ever developed, establishing a new benchmark for EFPI pressure-based systems. Numerous fabrication methods, including resin bonding, fusion splicing, and additive manufacturing, are investigated. In conjunction with this, multiple diaphragm reduction techniques such as manual polishing, automated polishing, and hydrofluoric acid etching are explored. The reason why we have not seen development of large core/diameter silica EFPI sensors, with advantages in sensitivity and resolution, is that the construction technique is difficult and unknown. The design construction, testing, and development of said large-diameter sensor is novel. This sub-Pascal resolution opens new possibilities for applications in microfluidics, atmospheric monitoring, and medical diagnostics where detecting minute pressure variations is critical. Finally, a comparative analysis of the sensor construction and diaphragm reduction methods provides insight into the future development of these high-performance EFPI sensors.

Saltmarshes are critical coastal ecosystems that are increasingly threatened by the accumulation of marine and terrestrial debris, including plastics, metals, wood, and other anthropogenic litter. This search focuses on a UAV based remote sensing framework for detecting and isolating such debris within saltmarsh environments by masking out natural background elements such as vegetation, soil, and water. High resolution multispectral was collected using a DJI Matrice 300 UAV equipped with an AGROWING Alpha 7R Sextuple camera and ground truthed via vegetation quadrats at Derrymore Island, Ireland. Spectral indices including NDVI, GNDVI, and MNDWI were employed to generate binary masks for key land cover types, enabling precise identification of anomalous debris. The proposed method demonstrates a scalable, noninvasive and semiautomated approach to debris detection in complex saltmarsh terrains. Results demonstrate high accuracy in debris detection and classification, highlighting UAV based multispectral imaging as an efficient, accurate, and scalable approach for environmental monitoring and marine debris assessment in sensitive coastal ecosystems.

Unmanned aerial vehicle (UAV) state estimation is fundamental across applications like robot navigation, autonomous driving, virtual reality (VR), and augmented reality (AR). This research highlights the critical role of robust state estimation in ensuring safe and efficient autonomous UAV navigation, particularly in challenging environments. We propose a deep learning-based adaptive sensor fusion framework for UAV state estimation, integrating multi-sensor data from stereo cameras, an IMU, two 3D LiDAR’s, and GPS. The framework dynamically adjusts fusion weights in real time using a long short-term memory (LSTM) model, enhancing robustness under diverse conditions such as illumination changes, structureless environments, degraded GPS signals, or complete signal loss where traditional single-sensor SLAM methods often fail. Validated on an in-house integrated UAV platform and evaluated against high-precision RTK ground truth, the algorithm incorporates deep learning-predicted fusion weights into an optimization-based odometry pipeline. The system delivers robust, consistent, and accurate state estimation, outperforming state-of-the-art techniques. Experimental results demonstrate its adaptability and effectiveness across challenging scenarios, showcasing significant advancements in UAV autonomy and reliability through the synergistic integration of deep learning and sensor fusion.

IoT applications are increasingly common, yet they often rely on expensive, externally managed authentication services. This paper introduces a novel, self-contained authentication method for IoT applications which leverages fog computing principles to lower operational costs and infrastructure complexity. The proposed system, fogauth, combines device serial numbers with cryptographically generated UUIDs to establish secure identification without third-party services. A static cloud-side architecture coupled with a lightweight, locally hosted API enables secure authentication through object-storage operations. Performance testing demonstrates comparable security performance to commercial cloud-based authentication while reducing long-term operational costs and maintaining latency at below 2 minutes in production conditions. fogauth provides a scalable and economically viable alternative for companies seeking to reduce cloud dependency and minimize long-term costs associated with IoT application security. To support reproducibility, a complete open-source implementation and validation dataset are provided, allowing independent replication and extension of the system.

Accurate classification of salt marsh vegetation is vital for conservation efforts and environmental monitoring, particularly given the critical role these ecosystems play as carbon sinks. Understanding and quantifying the extent and types of habitats present in Ireland is essential to support national biodiversity goals and climate action plans. Unmanned Aerial Vehicles (UAVs) equipped with optical sensors offer a powerful means of mapping vegetation in these areas. However, many current studies rely on single-sensor approaches, which can constrain the accuracy of classification and limit our understanding of complex habitat dynamics. This study evaluates the integration of Red-Green-Blue (RGB), Multispectral Imaging (MSI), and Hyperspectral Imaging (HSI) to improve species classification compared to using individual sensors. UAV surveys were conducted with RGB, MSI, and HSI sensors, and the collected data were classified using Random Forest (RF), Spectral Angle Mapper (SAM), and Support Vector Machine (SVM) algorithms. The classification performance was assessed using Overall Accuracy (OA), Kappa Coefficient (k), Producer’s Accuracy (PA), and User’s Accuracy (UA), for both individual sensor datasets and the fused dataset generated via band stacking. The multi-camera approach achieved a 97% classification accuracy, surpassing the highest accuracy obtained by a single sensor (HSI, 92%). This demonstrates that data fusion and band reduction techniques improve species differentiation, particularly for vegetation with overlapping spectral signatures. The results suggest that multi-sensor UAV systems offer a cost-effective and efficient approach to ecosystem monitoring, biodiversity assessment, and conservation planning.

This study presents a real-time, adaptive UAV system designed to enhance ecological surveys by overcoming the trade-off between wide-area coverage and high-resolution data collection. The Modular Detection and Targeting System (MDTS) integrates thermal imaging for broad detection and high-resolution RGB zoom imaging for precise species identification. Field trials demonstrated the system’s ability to detect and record both avian and mammalian species with significantly reduced redundant data and improved survey efficiency. Compared to traditional UAV methods, the MDTS achieved over 300-fold improvements in image resolution and up to a 1000-fold reduction in data volume. The system’s modular design enables rapid adaptation to diverse ecological applications, providing classification-ready data while minimizing post-processing demands. These results highlight the MDTS as a scalable, efficient tool for wildlife monitoring and environmental research, bridging the gap between detection and actionable ecological insights.

Saltmarshes are critical coastal ecosystems that offer biodiversity, shoreline protection, and high carbon sequestration. Monitoring these habitats is essential but challenging because of their spatial complexity and sensitivity. This paper presents a Unmanned Ariel Vehicle (UAV) based hyperspectral survey conducted over Derrymore Saltmarsh, County Kerry, Ireland, using a DJI Matrice 300 drone equipped with a Resonon Pika-L hyperspectral imager. We demonstrate the effectiveness of hyperspectral data and Spectral Angle Mapper (SAM) classification for identifying key saltmarsh vegetation species. Field data from vegetation quadrats were used for validation. Our results highlight the potential of UAV and HSI integration in supporting ecological research, improving habitat mapping accuracy, and enabling scalable, nondestructive monitoring.

This paper presents a comprehensive overview of cutting-edge autonomous forklifts, with a strong emphasis on sensors, object detection and system functionality. It aims to explore how this technology is evolving and where it is likely headed in both the near and long-term future, while also highlighting the latest developments in both academic research and industrial applications. Given the critical importance of object detection and recognition in machine vision and autonomous vehicles, this area receives particular attention. The article provides an in-depth summary of both commercial and prototype forklifts, discussing key aspects such as design features, capabilities and benefits, and offers a detailed technical comparison. Specifically, it clarifies that all available data pertains to commercially available forklifts. To obtain a better understanding of the current state-of-the-art and its limitations, the analysis also reviews commercially available autonomous forklifts. Finally, this paper includes a comprehensive bibliography of research findings in this field.

Unmanned Aerial Vehicles (UAV’s) State estimation is fundamental aspect across a wide range of applications, including robot navigation, autonomous driving, virtual reality, and augmented reality (AR). The proposed research emphasizes the vital role of robust state estimation in ensuring the safe navigation of autonomous UAVs. In this paper, we developed an optimization-based odometry state estimation framework that is compatible with multiple sensor setups. Our evaluation of the system is conducted using inhouse integrated UAV platform outfitted with multiple sensors including stereo cameras, an IMU, LiDAR sensors and GPS-RTK for ground truth comparison. The algorithm delivers robust and consistent UAV state estimation in various conditions including illumination changes, feature or structure-less environment or even during degraded Global Positioning System (GPS) signals or total signal loss, where single sensor SLAM mostly fails. The experimental findings demonstrate that the proposed method is superior in compare to current state-of-the-art techniques.

The ascent of electric vehicle (EV) technology as a leading solution for green transportation is accompanied by advancements in charging infrastructure and automation. A notable hindrance is the low level of automation in charging procedures. In response to this, Automatic Charging Robots (ACR) have emerged, equipped transitioning from the manual operation to an automated plugging and unplugging operation. However, for this process to be executed flawlessly, these robots necessitate a charging port detection system with a precise navigation system to ensure accurate insertion of the charging gun into the designated charging port. This paper presents a sophisticated system, AViTRoN (Advanced Vision Track Routing and Navigation), which is developed for Automated Charging Robots in the context of Electric Vehicle (EV) charging. AViTRoN integrates advanced technologies to enable efficient charging port detection, navigation, and seamless user interaction. Utilizing the YOLOv8 deep learning model, AViTRoN ensures real-time charging port type detection using the data from a 3D depth sensor and an IR sensor within the Robot Operating System (ROS) framework. The 3D depth sensor provides detailed spatial information, while the IR sensor detects subtle environmental changes, enhancing the system’s accuracy during operation. AViTRoN also incorporates a charging completion notification mechanism, sending instant alerts to users via GSM/GPRS communication upon the conclusion of the charging cycle, thereby enhancing user convenience and experience.

Remote locations including but not limited to, far-reaching offshore airspaces, provide limited communications capabilities between equipment and ground stations. The case for DAVs is no exception, all the while requiring swift decisions because of the high speeds, lack of detailed real-time weather data, and complex applications. With the limitations and require-ments at hand, we propose real-time edge fault-tolerant SCADA (detection of anomalies) based on learning continuous sequential data for efficient control of the UAV's altitude and heading. Our data-driven solution, involving an extensive suite of sensory data processing, demonstrates the potential to significantly reduce communication and decision-making capabilities in cases of remote locations, enabling safer and more efficient UAV operations. The proposed system leverages machine learning algorithms to analyze real-time data from both extrinsic and intrinsic UAV sensors, allowing for predictive control and fault detection. By processing data at the edge, our solution reduces the need for bandwidth-intensive data transmission minimizing latency and ensuring swift and reliable decision-making. We show the data collection and training in a high-fidelity aviation simulator that closely matches real flight conditions.