Search Results (135)

Multi-agent perception systems must operate under fundamental asymmetries: some agents provide fast but unreliable observations, while others deliver higher-quality evidence with delay and uncertain correspondence. Traditional deterministic orchestration and rule-based fusion struggle to manage these trade-offs, often producing brittle or unstable behavior. We introduce a probabilistic orchestration framework that treats coordination as an epistemic generation problem—constructing and updating belief states under uncertainty—rather than a selection problem. Instead of committing to a single agent’s output, the orchestrator constructs a belief state that explicitly represents uncertainty, evidential provenance, and temporal relevance. Decisions are produced through latency-aware, association-weighted fusion, and uncertainty itself becomes a first-class signal governing action, deferral, and learning. Crucially, the orchestrator enables controlled teacher–student adaptation: high-confidence, well-associated stationary observations are gated into a feedback loop that improves ego perception over time while mitigating error amplification. We demonstrate the approach on an infrastructure-assisted dual-camera obstacle-recognition task. Experimental results show improved robustness to distance, occlusion, and delayed evidence compared to ego-only and deterministic orchestration baselines. By operationalizing orchestration as epistemic generation, this work provides a unifying framework for robust decision-making and safe adaptation in multi-agent systems, with implications that extend beyond perception to agentic and generative AI architectures. Full article

The Internet of Things (IoT) uses networked devices, dispersed sensors, and cameras to create huge, diverse data streams. Concept drift, in which the underlying data distribution shifts over time, is frequently caused by the non-stationary and multimodal character of these streams. Static machine learning models, based on fixed data distributions, reduce forecast accuracy and system reliability since they are unable to adapt to such changes. This paper proposes an Adaptive Multimodal Long Short-Term Memory (AM-LSTM) architecture to address these challenges by combining modality-specific temporal modelling, attention-based dynamic fusion, and drift-aware online learning. An attention mechanism adaptively weights informative streams to mitigate the impact of noisy or missing input, while specialist LSTM encoders capture the temporal correlations of each modality. Concept drift is detected using a sliding-window error monitoring technique, and adaptive learning rate adjustment and selective retraining are started when significant distributional changes occur. The proposed system is tested under synthetic drift conditions using the Edge-IoT and UNSW-NB15 benchmark datasets. Experimental results demonstrate that AM-LSTM achieves 88.7% accuracy and an F1-score of 0.85, adapting to drift within 620 samples while maintaining an average update latency of 47 ms per batch. Compared with static and existing adaptive baselines, the proposed approach provides improved robustness, faster drift adaptation, and computational efficiency suitable for real-time IoT environments. Full article

Remote photoplethysmography (rPPG) enables contactless estimation of cardiovascular signals from video, but most existing studies assume a fixed, stationary camera. This study investigates the feasibility of performing real-time moving-average rPPG (MA-rPPG) using a drone-mounted camera, where platform motion, vibration, and viewing distance introduce additional challenges. Building on our previously validated real-time MA-rPPG smart mirror platform, we reuse the smart mirror interface as a unified frontend for visualization, synchronization, and logging while adapting the MA-rPPG pipeline to operate on live video streamed from an off-the-shelf DJI Tello micro-drone. Feasibility experiments were conducted with 10 participants under controlled indoor lighting and constrained flight conditions, where the drone maintained a stable hover in front of a standing subject and facial video was processed in real time to estimate heart rate from a forehead region of interest. To avoid cross-modality bias and clarify the effect of the aerial imaging platform, drone-derived MA-rPPG outputs were compared against a fixed desktop-camera MA-rPPG reference using the same trained model, enabling a controlled, like-for-like evaluation. The results indicate that continuous heart-rate estimation from a drone camera is feasible in our controlled hover-only setup, while agreement tended to vary with hover stability and effective facial resolution. This work is presented strictly as a feasibility-stage investigation and does not claim clinical validity. The findings provide an experimental baseline and operating-envelope insight for future motion-robust rPPG on mobile and aerial health-sensing platforms. Full article

Mid-sized Philippine cities commonly rely on fixed-time traffic signal plans that cannot respond to short-term, demand-driven surges, resulting in measurable idle time at stop lines, increased delay, and unnecessary emissions, while adaptive signal control has demonstrated performance benefits, many existing solutions depend on centralized infrastructure and high-bandwidth connectivity, limiting their applicability for resource-constrained local government units (LGUs). This study reports a field deployment of TrafficEZ, a lightweight edge AI signal controller that reallocates green splits locally using traffic-density approximations derived from cabinet-mounted cameras. The controller follows a macroscopic, cycle-level control abstraction consistent with Transportation System Models (TSMs) and does not rely on stationary flow–density–speed (fundamental diagram) assumptions. The system estimates queued demand and discharge efficiency on-device and updates green time each cycle without altering cycle length, intergreen intervals, or pedestrian safety timings. A quasi-experimental pre–post evaluation was conducted at three signalized intersections in El Salvador City using an existing 125 s, three-phase fixed-time plan as the baseline. Observed field results show average per-vehicle delay reductions of 18–32%, with reclaimed effective green translating into approximately 50–200 additional vehicles per hour served at the busiest approaches. Box-occupancy durations shortened, indicating reduced spillback risk, while conservative idle-time estimates imply corresponding CO₂ savings during peak periods. Because all decisions run locally within the signal cabinet, operation remained robust during backhaul interruptions and supported incremental, intersection-by-intersection deployment; per-cycle actions were logged to support auditability and governance reporting. These findings demonstrate that density-driven edge AI can deliver practical mobility, reliability, and sustainability gains for LGUs while supporting evidence-based governance and performance reporting. Full article

Photogrammetric reconstruction from indoor imagery requires either labor-intensive ground control points (GCPs) or positioning sensor integration. While global navigation satellite system technology revolutionized aerial photogrammetry by enabling direct georeferencing through integrated sensor orientation (ISO), indoor environments lack an equivalent positioning solution. Before indoor positioning systems can be adopted for photogrammetric applications, their fundamental measurement precision must be established. This study characterizes the repeatability and temporal stability of the ZeroKey Quantum real-time location system (RTLS) as a prerequisite to testing reconstruction accuracy when RTLS measurements provide camera pose constraints in photogrammetric bundle adjustment. Through systematic tripod-mounted observations across 30 test locations in a controlled laboratory environment, optimal data collection protocols were determined, temporal stability was investigated, and measurement precision was quantified. An automated position-based stationary detection algorithm using a 20 mm threshold successfully identified all 30 stationary periods for durations of 30 s or less. Optimal duration analysis revealed that 1 s observation windows achieve 3 mm position precision and 1° orientation precision after brief settling, enabling practical workflows with worst-case total collection time of 2.5 s per station. Per-axis uncertainties were quantified as 1.6 mm, 1.7 mm, and 1.1 mm root mean square (RMS) for position and 0.08°, 0.09°, and 0.07° RMS for orientation. These findings demonstrate that ultrasonic RTLS achieves millimeter-level position repeatability and sub-degree orientation repeatability, establishing the measurement precision necessary to justify subsequent accuracy testing through photogrammetric bundle adjustment. Full article

An ultra-wideband (UWB) Anchor–Tag commercial sensor system used for positioning is characterized herein, against an image-processing based positioning system used as a ground truth. The UWB consists of a single anchor that measures the angle of arrival (AoA) and distance to the moving tag. The driftless camera-based positioning system requires a series of complex operations, among camera calibration, image processing and network transmission delay estimation, and time alignment with the analyzed UWB measurement system. For the UWB system, the accuracy, precision, resolution, covered area, and error-vs-distance dependence are measured on several collected trajectories, both stationary and in motion. Several filtering solutions are proposed to improve these metrics that are affected by some faulty measurements, to subsequently validate the overall performance. The condition monitoring is verified both in offline and in online processing modes, using these filtering solutions. Our approach is black-box and does not use additional information except for raw position data. The importance and feasibility of UWB systems for indoor or outdoor localization is demonstrated, as well as some caveats and possible mitigation strategies. Full article

Modern civil engineering structures, such as footbridges, are increasingly susceptible to vibrations induced by human activities, emphasizing the importance of accurately assessing crowd-induced loading. Developing realistic load models requires detailed insight into the underlying crowd dynamics, which in turn depend on the coordination between individuals and the spatial organization of the group. A deeper understanding of these human–human interactions is therefore essential for capturing the collective behaviour that governs crowd-induced vibrations. This paper presents a vision-based trajectory reconstruction methodology that captures individual movement trajectories in both small groups and large-scale running events. The approach integrates colour-based image segmentation for instrumented participants, deep learning–based object detection for uninstrumented crowds, and a homography-based projection method to map image coordinates to world space. The methodology is applied to empirical data from two urban running events and controlled experiments, including both stationary and dynamic camera perspectives. Results show that the framework reliably reconstructs individual trajectories under varied field conditions, applicable to both walking and running activities. The approach enables scalable monitoring of human activities and provides high-resolution spatio-temporal data for studying human–human interactions and modelling crowd dynamics. In this way, the findings highlight the potential of vision-based methods as practical, non-intrusive tools for analysing human-induced loading in both research and applied engineering contexts. Full article

Remote photoplethysmography (rPPG) is a non-contact technology that estimates physiological signals, such as Heart Rate (HR), by capturing subtle skin color changes caused by periodic blood volume variations using only a standard RGB camera. While cost-effective and convenient, it suffers from a fundamental limitation: performance degrades severely in dynamic environments due to susceptibility to noise, such as abrupt illumination changes or motion blur. This study presents a deep learning framework that combines two structural modifications to ensure robustness in dynamic environments, specifically modeling movement noise and illumination change noise. The proposed framework structurally cancels global disturbances, such as illumination changes or global motion, through a dual-branch pipeline that encodes the face and background in parallel after Video Color Magnification (VCM) and then performs differencing. Subsequently, it utilizes a structure that injects a Temporal Shift Module (TSM) into the Spatio-Temporal Feature Extraction (SSFE) block to preserve long- and short-term temporal correlations and smooth noise, even amidst short and irregular movements. We measured MAE, RMSE, and correlation on the standard dataset UBFC-rPPG under four noise conditions: clean, illumination change noise, Movement Noise, Both Noise and the real-world in-vehicle dataset MR-NIRP (Stationary and Driving). Experimental results showed that the proposed method achieved consistent error reduction and correlation improvement compared to the VS-Net baseline in the illumination change noise-only and combined noise environments (UBFC-rPPG) and in the high-noise driving scenario (MR-NIRP). It maintained competitive performance in motion-only noise. Conversely, a modest performance disadvantage was observed under clean conditions (UBFC) and quasi-clean stationary conditions (MR-NIRP), interpreted as a design trade-off focused on global noise cancellation and temporal smoothing. Ablation studies demonstrated that the dual-branch pipeline is the primary contributor under illumination change noise, while TSM is the key contributor under movement noise, and that the combination of both elements achieves optimal robustness in the most complex scenarios. Full article

Recent progress in computer vision and embedded systems has facilitated real-time monitoring of bioprocesses; however, lightweight and scalable solutions for resource-constrained settings remain limited. This work presents a modular framework for monitoring Chlorella vulgaris growth by integrating RGB image processing with multimodal sensor fusion. The system incorporates a Logitech C920 camera and low-cost pH and temperature sensors within a compact photobioreactor. It extracts RGB channel statistics, luminance, and environmental data to generate a 10-dimensional feature vector. A feedforward artificial neural network (ANN) with ReLU activations, dropout layers, and SMOTE-based data balancing was trained to classify growth phases: lag, exponential, and stationary. The optimized model, quantized to 8 bits, was deployed on an ESP32 microcontroller, achieving 98.62% accuracy with 4.8 ms inference time and a 13.48 kB memory footprint. Robustness analysis confirmed tolerance to geometric transformations, though variable lighting reduced performance. Principal component analysis (PCA) retained 95% variance, supporting the discriminative power of the features. The proposed system outperformed previous vision-only methods, demonstrating the advantages of multimodal fusion for early detection. Limitations include sensitivity to lighting and validation limited to a single species. Future directions include incorporating active lighting control and extending the model to multi-species classification for broader applicability. Full article

Ports with Mediterranean-like traffic profiles combine dense passenger, cargo, touristic, and local operations in confined waters where many small craft sail without AIS, increasing collision risk. Nature of such traffic in often unpredictable, due to often and sudden course corrections or changes. In such situations, it is possible that larger ships cannot manoeuvre to avoid collisions with small vessels. Hence, it is important to the port authority to develop a fast and adoptable mean to reduce collision risks. We present an end-to-end shore-based framework that detects and tracks vessels from fixed cameras (YOLOv9 + DeepSORT), estimates speed from monocular lateral video with an artificial neural network (ANN), and visualises collision risk in augmented reality (AR) for VTS/port operators. Validation in the Port of Split using laser rangefinder/GPS ground truth yields MAE 1.98 km/h and RMSE 2.18 km/h (0.605 m/s), with relative errors 2.83–21.97% across vessel classes. We discuss limitations (sample size, weather), failure modes, and deployment pathways. The application uses stationary port camera as an input. The core calculations are performed at user’s computer in the building. Mobile application uses wireless communication to show risk assessment at augmented reality smart phone. For training of ANN, we used The Split Port Ship Classification Dataset. Full article

Playgrounds are global environments that are purpose made for children and can offer a variety of opportunities for children to be physically active and practice their fundamental movement skills (FMS), which can lead to future physical activity and sport participation. Previous research highlighted that children engage in different types of physical activity (PA) depending on playgrounds apparatus and area. However, there is a paucity of research that investigates the link between playground features, structures, PA, and FMS. This study sought to assess the impact of different playgrounds on PA type PA intensity and the types of FMS completed. This observational study examined 29 (M = 10, F = 19) children’s behaviours on three different playgrounds. Video cameras were placed strategically across the three playgrounds to allow for footage to be captured and analysed using the Observational System for Recording Physical Activity in Children (OSRAC). One-way ANOVA was used to examine the different OSRAC categories across the three playgrounds. Climbing equipment (average 1217.10 s) was the frequently used type of apparatus, standing was the most commonly performed type of activity (average 377.60 s) and stationary movements whilst moving limbs were the most regularly (average 605.13 s) performed type of PA intensity. There were no instances of any throwing, catching, or kicking activities performed across the three playgrounds. Results suggest that public playgrounds do not facilitate more intense types of PA, nor object control skills due to a lack of suitable equipment. Full article

In ports with a significant density of non-AIS vessels, there is an increased risk of collisions. This is because physical limitations restrict the maneuverability of AIS vessels, while small vessels that do not have AIS are unpredictable. To help with collision prevention, we propose an augmented reality system that detects vessels from video stream and estimates speed with a single sideway-mounted camera. The goal is to visualize a cone for risk assessment. The estimation of speed is executed by geometric relations between the camera and the ship, which were used to estimate distances between points in a known time interval. The most important part of the proposal is vessel speed estimation by a monocular camera validated by a laser speed measurement. This will help port authorities to manage risks. This system differs from similar trials as it uses a single stationary camera linked to the authorities and not to the bridge crew. Full article

The main function of triangulation-based laser profile sensors—also referred to as laser profilometers or profilers—is the three-dimensional scanning of moving objects using laser triangulation. In addition to capturing 3D data, these profilometers simultaneously generate grayscale images of the scanned objects. However, the quality of these images is often degraded due to interference of the laser light, manifesting as speckle noise. In profilometer images, this noise typically appears as vertical stripes. Unlike the column fixed pattern noise commonly observed in TDI CMOS cameras, the positions of these stripes are not stationary. Consequently, conventional algorithms for removing fixed pattern noise yield unsatisfactory results when applied to profilometer images. In this article, we propose an effective method for suppressing speckle noise in profilometer images of flat surfaces, based on local column median vectors. The method was evaluated across a variety of surface types and compared against existing approaches using several metrics, including the standard deviation of the column mean vector (SDCMV), frequency spectrum analysis, and standard image quality assessment measures. Our results demonstrate a substantial improvement in reducing column speckle noise: the SDCMV value achieved with our method is 2.5 to 5 times lower than that obtained using global column median values, and the root mean square (RMS) of the frequency spectrum in the noise-relevant region is reduced by nearly an order of magnitude. General image quality metrics also indicate moderate enhancement: peak signal-to-noise ratio (PSNR) increased by 2.12 dB, and the structural similarity index (SSIM) improved from 0.929 to 0.953. The primary limitation of the proposed method is its applicability only to flat surfaces. Nonetheless, we successfully implemented it in an optical inspection system for the furniture industry, where the post-processed image quality was sufficient to detect surface defects as small as 0.1 mm. Full article

The elevator light curtain is constrained by technical limitations in its infrared detection mechanism; thus, it is difficult to effectively identify the transparent material and elongated form of the object, which has become one of the main causes of abnormal elevator jamming accidents. To mitigate elevator accidents, we propose a novel visual segmentation method, PKNet (Prior Knowledge Network), specifically designed for detecting transparent and slender objects. We observe that the majority of cameras used in elevators are stationary, resulting in an inherently static background, while vision tasks primarily focus on detecting foreground objects. To this end, PKNet enhances the segmentation of dynamic foreground objects by incorporating prior knowledge of the static background and the characteristics of foreground objects. We also introduce ETAS-D, the first dataset designed for the segmentation of transparent and slender anomalous objects in elevator environments. This dataset consists of 4797 image frames, each with meticulously annotated masks of transparent and slender objects, captured from multiple viewpoints of 10 elevators. Extensive experimental results demonstrate that PKNet significantly outperforms existing methods in this domain. Furthermore, we propose a synthetic data generation workflow specifically designed for slender objects to enhance the model’s generalization ability and reliability. Full article

We developed a multi-feature drowsiness detection model employing eye aspect ratio (EAR), mouth aspect ratio (MAR), head pose angles (yaw, pitch, and roll), and a Raspberry Pi 4 for real-time applications. The model was trained on the NTHU-DDD dataset and optimized using long short-term memory (LSTM) deep learning algorithms implemented using TensorFlow version 2.14.0. The model enabled robust drowsiness detection at a rate of 10 frames per second (FPS). The system embedded with the model was constructed for live image capture. The camera placement was adjusted for optimal positioning in the system. Various features were determined under diverse conditions (day, night, and with and without glasses). After training, the model showed an accuracy of 95.23%, while the accuracy ranged from 91.81 to 95.82% in validation. In stationary and moving vehicles, the detection accuracy ranged between 51.85 and 85.71%. Single-feature configurations exhibited an accuracy of 51.85 to 72.22%, while in dual features, the accuracy ranged from 66.67 to 75%. An accuracy of 80.95 to 85.71% was attained with the integration of all features. Challenges in the drowsiness included diminished accuracy with MAR alone and delayed prediction during transitions from non-drowsy to drowsy status. These findings underscore the model’s applicability in detecting drowsiness while highlighting the necessity for refinement. Through algorithm optimization, dataset expansion, and the integration of additional features and feedback mechanisms, the model can be improved in terms of performance and reliability. Full article