Search Results (4,427)

Mobile robots used in edge-assisted Industrial Internet-of-Things (IIoT) settings generate coupled motion, LiDAR, edge-compute, and network telemetry. Public datasets that place these streams in one tabular format, with scenario labels suitable for machine-learning experiments, are still limited. This data descriptor presents a pilot-real-calibrated indoor robotic IoT benchmark dataset with 120,000 records sampled at 2 Hz across nominal navigation and nine anomaly scenarios. The benchmark rows are generated from physically constrained simulation rules and are explicitly labeled as synthetic benchmark data. Real pilot evidence is included separately: ROS Noetic runs on a TurtleBot3 Burger, successful LD08 LiDAR bringup after resolving a driver mismatch, and NVIDIA Jetson Nano tegrastats logs under normal-navigation workloads. The calibrated file aligns normal-navigation LiDAR and edge-compute distributions with these pilot measurements while keeping the multi-scenario structure needed for controlled anomaly-detection experiments. The package includes CSV files, metadata, a data dictionary, validation reports, baseline scripts, ROS collection utilities, and a plan for future fully physical data collection. The complete dataset is openly available on Zenodo. Full article

To achieve non-contact and continuous monitoring of body length in Micropterus salmoides and overcome the stress damage and subjective error associated with traditional manual measurement, this paper proposes an improved YOLOv8-based multi-target tracking framework for intensive recirculating aquaculture systems. The system employs a geometric measurement framework based on monocular vision that achieves conversion from pixel coordinates to actual body length through camera calibration, water-surface refraction correction, and pose projection correction. Under a collaborative optimization framework integrating detection and tracking, the model incorporates multi-scale feature enhancement, lightweight re-identification (ReID), and a robust data association mechanism, which improves system stability under conditions of high fish density, variable illumination, and turbid water. A shallow feature fusion path is introduced to enhance small-target perception, and a MobileNetV3_ReID model is adopted to extract highly discriminative appearance features, which improves identity consistency while maintaining model compactness. In the data association stage, a hybrid cost matrix integrating IoU, cosine similarity, and motion consistency is constructed, and optimal matching is realized through the Hungarian algorithm. Dynamic threshold adjustment and an exponential moving-average feature-update strategy are introduced to effectively suppress identity switching. Experiments were conducted on an overhead video dataset of Micropterus salmoides collected at a recirculating aquaculture system facility. The results show that the proposed method achieves 82.7% mAP50 while maintaining a real-time throughput of 88 FPS, with MOTA reaching 76.9% and IDF1 reaching 81.5%—the latter representing an improvement of 3.2 percentage points over BoT-SORT and 5.3 percentage points over the YOLOv8 baseline tracker. The number of identity switches (IDSW) decreased from 89 in the baseline configuration to 39, a reduction of 56.2%. Crucially, these component-level improvements translate into a body length error (BLE) of 5.2 ± 1.8% (MAE = 1.35 cm, Pearson r = 0.972), representing a 38.8% improvement over the baseline BLE of 8.5% and satisfying the 5–10% tolerance required for aquaculture growth monitoring. Ablation analysis confirms that both detection enhancements (contributing −1.3% BLE) and tracking optimizations (contributing −2.0% BLE) are necessary to achieve this application-level accuracy. Full article

Background/Objectives: Topical imiquimod is an established non-invasive treatment for superficial basal cell carcinoma (sBCC). However, data on treatment-associated changes in tumor microvascularization remain limited. This study investigated vascular changes before and after imiquimod therapy using multimodal non-invasive imaging. Methods: In this single-center, prospective observational study, 31 basal cell carcinomas in 20 patients were examined before and 12–16 weeks after topical imiquimod therapy (5%, five times weekly for six weeks) using dermoscopy, dynamic optical coherence tomography (D-OCT), and line-field confocal optical coherence tomography (LC-OCT). Analyses were performed as paired before-and-after comparisons. While approved for sBCC, a small number of thin nodular and infiltrative BCCs were included exploratorily; subgroup analyses were not powered. Results: Dermoscopy showed a nominally significant shift toward smaller vessel diameter categories after therapy (ATS = 8.183, df = 1, p = 0.004). D-OCT-derived parameters (vessel density, vessel diameter, and depth of the vascular plexus) did not show nominally significant changes. LC-OCT showed nominally lower apparent intratumoral flow scores (ATS = 13.285, df = 1, p < 0.001), reduced occurrence of vessel-wall-associated intraluminal structures showing a rolling-like motion pattern (86.7% before treatment versus 33.3% after treatment; ATS = 13.357; df = 1, p < 0.001), and a reduction in maximum vessel diameter (ATS = 6.110, df = 1, p = 0.013). The primary LC-OCT inferential analyses were performed at the lesion level without adjustment for within-patient clustering and should therefore be interpreted as exploratory. An additional patient-cluster-adjusted paired change-score sensitivity analysis for LC-OCT maximum vessel diameter yielded a directionally consistent estimate (−17.81 µm; 95% CI: −34.40 to −1.23; p = 0.037). The primary exploratory endpoints were LC-OCT–based apparent intratumoral flow and maximum vessel diameter; secondary endpoints included dermoscopic and D-OCT–based vascular parameters. In the exploratory response-stratified analysis, the change in LC-OCT-based maximum vessel diameter did not differ significantly among the assigned response groups (Kruskal–Wallis H = 3.870, df = 2, raw p = 0.144; BH-adjusted p = 0.753). Conclusions: LC-OCT detected several exploratory vascular changes between the pre-treatment examination and follow-up and may provide complementary information for the non-invasive assessment of BCC after imiquimod therapy. Given the exploratory design, limited sample size, and lack of systematic histological confirmation, these findings are hypothesis-generating and require validation in larger prospective studies. Full article

Airport turnaround is an important operational process that directly affects flight punctuality, airport capacity, and ground-handling efficiency. However, many turnaround activities are still monitored manually or through fragmented operational records, which can limit real-time visibility and delay identification. This study proposes a computer vision-based airport turnaround monitoring pipeline that integrates YOLOv11 object detection, Norfair multi-object tracking, and frame differencing-based motion analysis to extract key operational events from airport video footage. Publicly available turnaround footage from Shinshu Matsumoto Airport, Japan, was collected under different environmental conditions, including daytime, nighttime, rainy, after-rain, and transition lighting conditions. From selected videos, 1446 images were labeled into 11 airport turnaround object classes, including tow tug, aerobridge, airplane, baggage container, belt loader, belt loader roof, fuel line, fuel tanker, fuel tube, tractor, and window. The dataset was divided into training, validation, and testing sets using a 70:20:10 ratio. The trained YOLOv11 model achieved strong detection performance, with overall test an precision of 0.9609, recall of 0.9445, and mAP50 of 0.9617. To support activity-level interpretation beyond object detection, the proposed pipeline applies frame differencing within specific regions of interest, including the aerobridge window region for passenger deboarding and boarding detection, and the belt loader roof region for baggage unloading and loading detection. The extracted object detections, motion spikes, and temporal logs are then converted into a Gantt chart that summarizes major turnaround activities, including airplane parking, deboarding, baggage unloading, refueling, baggage loading, boarding, and pushback. The results demonstrate that the proposed modified YOLO-based pipeline can transform ordinary airport video footage into structured operational timelines, supporting more transparent, data-driven, and automated monitoring of airport turnaround processes. Full article

Traditional biomechanical analysis is constrained by high capital costs and the physical limitations imposed by markers, posing significant barriers to clinical adoption. This review evaluates the emergence of artificial intelligence (AI)-based markerless motion capture (MMC) as a transformative approach for democratizing movement science in clinical rehabilitation. The discussion outlines the progression from legacy geometric visual hulls to advanced deep learning architectures, with particular focus on YOLO-based two-dimensional detection and spatio-temporal transformer models for three-dimensional pose estimation. Evidence indicates that multi-camera MMC frameworks achieve research-grade positional accuracy (16–34 mm Mean Per-Joint Position Error—MPJPE), while monocular systems provide sufficient sensitivity (82–88%) for longitudinal monitoring of geriatric fall risk and stroke recovery. While challenges persist in achieving precise axial rotation measurement, integrating real-time signal refinement enables objective and ecologically valid assessments in community-based healthcare settings. This technological advancement redefines movement analysis, shifting it from a laboratory-bound procedure to a widely accessible and interoperable diagnostic tool. Full article

Background subtraction (BS) is a fundamental technique in intelligent transport vision systems, widely used to detect and track moving objects, such as vehicles, pedestrians and obstacles, in driving environments. It plays a crucial role in advanced driver-assistance systems (ADAS) and autonomous driving by enabling scene understanding and real-time motion analysis. However, BS processing must be optimized when targeting real-time processing on resource-constrained embedded systems, which present significant challenges due to limited computational power, memory constraints, and strict real-time requirements. Among the most commonly used BS techniques, the Codebook model and Gaussian Mixture Models (GMM) are known for their higher accuracy and light-model compared to many deep learning-based BS. In this work, we propose a fully heterogeneous CPU and GPU parallel implementation of both Codebook and GMM algorithms with an auto-load balancing over the processing units. This approach has been evaluated on the low-cost Jetson Orin Nano platform from NVIDIA, enabling efficient workload balancing across heterogeneous hardware resources. The suggested solution yields significant performance improvements over the state-of-the-art, achieving 59 frames per second (FPS) for GMM and 66 FPS for the Codebook method on full-HD (1080p) video streams. The results confirm the effectiveness of the proposed method in accelerating BS and demonstrate its suitability for real-time deployment in resource-constrained embedded environments. Full article

Infrared UAV detection plays a crucial role in both security surveillance and military applications. However, under fast UAV movement or dynamic zooming scenarios, the rapid scale variation of targets poses severe challenges to existing detection models, especially on resource-constrained edge devices. To address this, a lightweight scale-adaptive multi-scale feature fusion model, termed LMF-IR, is proposed for efficient and accurate detection under sudden target size changes. The model integrates three key components: a Multi-Dilation Residual Block (MDRB) for enhanced multi-scale feature representation, an improved Channel Attention Model–Feature Fusion Pyramid Network (CAM-FPN) to boost adaptive feature fusion, and a modified P-WIoU loss function designed for precise bounding box regression under varying target sizes. The MDRB module effectively captures fine-grained features across multiple scales and reliably identifies targets of varying sizes. The CAM-FPN incorporates a channel attention mechanism, which can dynamically adjust the weights of features, enabling the model to focus on informative feature channels. The redesigned P-WIoU loss function is designed to account for the shape characteristics of UAV target bounding boxes. It includes centroid distance, overlap ratio, and aspect ratio, thereby improving localization accuracy under rapid scale changes. The experimental results on our self-built UAV–infrared dataset show that LMF-IR reduces 1.4 G in floating-point operations compared to the baseline model, and the parameter count is reduced to 62% of the baseline. At the same time, mAP@0.5:0.95 increases by 2.4%. Moreover, on the public ANTI-UAV dataset, our method increases mAP@0.5:0.95 by 4.8%, indicating that our method has excellent performance in real-time infrared UAV detection under rapid target scale changes. Full article

Deep learning perception in autonomous vehicles (AVs) has created a critical attack surface in which adversarial patches and sensor-spoofing perturbations cascade from perception errors into unsafe driving decisions. Existing defenses face three limitations: most require retraining the perception network, making them impractical for already-deployed fleets; they operate almost exclusively at the perception layer, without verifying whether a compromised detection actually altered the driving action; and they leave temporal consistency across frames largely unexploited. This paper presents SENTINEL, a zero-modification, plug-and-play defense that wraps any deployed AV perception-and-planning stack without updating its weights, calibrating only the detection thresholds, score combination weights, and reference exemplars once on a small held-out calibration set. SENTINEL integrates a frozen foundation model verification ensemble (CLIP, DINOv2, SAM-2), a temporal consistency scorer that flags patches through anomalous frame-to-frame stability under ego-motion, a counterfactual policy verifier that replans under reconstructed perception and measures action-space divergence, and a risk-adaptive safety shield that modulates driving aggressiveness by verification confidence. Across CARLA, nuScenes, KITTI, and BDD100K, against five adversarial attacks and an adaptive adversary, SENTINEL reduces the attack success rate by up to 92%, keeps the clean accuracy loss to approximately 1.8 percentage points, reduces the collision rate under attack by approximately 87%, and adds under 45 ms latency on an RTX 4090 GPU. SENTINEL reframes adversarial robustness as a runtime property of the complete autonomous decision pipeline. Full article

Mobile inspection robots require reliable license plate detection and recognition under constrained computing resources, small-scale or distant imaging conditions, motion blur, and complex background interference. To address these coupled challenges, this paper proposes a lightweight robot-view visual sensing framework for CPU-oriented license plate perception. Instead of simply stacking network modules, the proposed framework follows a unified design principle of reducing redundant computation while compensating for task-critical visual information. In the detection stage, a YOLOv8-MGL detector is developed based on YOLOv8n by combining GhostC2f-based lightweight feature aggregation with LSKAlite-based contextual enhancement after the SPPF module. In the recognition stage, SimAM is embedded into LPRNet to enhance discriminative character responses under motion blur, low resolution, and local degradation without introducing additional learnable parameters. Experiments on the held-out EDRV-LP test set show that YOLOv8-MGL achieves 99.5% mAP₅₀ and 71.1% mAP_50:95, while reducing the number of parameters from 3.01 M to 2.77 M and GFLOPs from 8.1 to 7.5 compared with YOLOv8n. On a CPU-only Intel Xeon Platinum 8260C platform, YOLOv8-MGL achieves 23.98 FPS. SimAM-LPRNet improves the module-level cropped-plate recognition accuracy from 83.10% to 87.17%. To further examine system-level feasibility, a supplementary YOLOv8-MGL + CRNN-CTC pipeline is evaluated from raw images to final plate strings, achieving 91.0% exact recognition accuracy on the held-out EDRV-LP test set, 92.0% on a non-overlapping external CCPD subset, and 13.25 FPS for complete CPU-only processing. These results demonstrate that the proposed framework provides a favorable trade-off among model compactness, localization quality, recognition robustness, and CPU-oriented inference feasibility for mobile robotic inspection scenarios. Full article

Product recognition in e-commerce live streaming is hindered by rapid viewpoint changes, occlusions, motion blur, and inconsistencies between visual and spoken information. Existing approaches typically focus on individual components such as detection, OCR, or speech recognition, which limits their effectiveness in end-to-end structured product understanding. To address this problem, we propose an integrated framework that combines task-oriented keyframe selection with multimodal semantic fusion. The framework first uses D-FINE to localize product regions and then selects informative frames through two complementary strategies. Strategy A considers both detection confidence and Laplacian-based sharpness, while Strategy B combines detection confidence with a learned quality component estimated by an EfficientNetV2-M regression model. OCR, visual-semantic recognition, and ASR are then applied to extract complementary evidence, and a Qwen3.5-27B large language model is used to structure and fuse multimodal evidence into standardized product outputs, including brand, product name, and category. Experiments on an in-house e-commerce livestreaming dataset demonstrate substantial gains over a last-frame baseline. Strategy B achieves the best overall result, improving the Perfect Match Rate from 0.609 to 0.775 and the Semantic Similarity from 0.697 to 0.802. Ablation studies further show that the full multimodal framework consistently outperforms unimodal and dual-modality variants under both frame selection strategies. In addition, Top-K analysis indicates that single-frame inference provides a practical balance between OCR evidence completeness and efficiency. Efficiency analysis shows that the per-video API monetary cost remains low under the pricing configuration used in this study, while API latency is mainly limited by Qwen3.5-27B LLM calls for evidence structuring and final fusion. Overall, the proposed framework offers an effective and extensible solution for structured product recognition in complex live-streaming scenarios. Full article

Early smoke detection for forest fire prevention requires accurate and temporally stable decisions under dynamic clutter, tiny long-range targets, atmospheric degradation, and partial sensor unreliability. This paper presents FS-YOLOv3, a reliability-driven RGB–thermal smoke detector that extends a reproduced FS-YOLO baseline with two new modules: Cross-Temporal Consistency Alignment (CTCA) and Scene-Adaptive Expert Routing Fusion (SAERF). CTCA performs local short-horizon feature alignment and is evaluated with additional offset-field diagnostics to test whether the learned offsets correlate more strongly with annotated smoke expansion than with non-smoke motion. SAERF routes fused features to compact experts according to illumination, haze, texture ambiguity, and thermal reliability, with descriptor ablations and collinearity diagnostics used to examine routing stability. On the proposed clip-level RGB–thermal benchmark, FS-YOLOv3 improves over the reproduced FS-YOLO baseline from 93.7% to 96.3% mAP@0.5 and from 89.5% to 94.8% temporal alarm consistency (TAC), with 165 model FPS on Jetson AGX Orin under the default one-frame-look-ahead buffered inference setting. Comparisons with lightweight YOLO detectors, RGB-only and infrared-only controls, simple fusion strategies, and stronger temporal baselines provide deployment context, while the main technical evidence is the controlled gain obtained by enabling CTCA and SAERF on the same baseline architecture. To support reproducibility, the paper specifies the baseline interface, sensor and annotation protocol, sequence-disjoint split policy, temporal metrics, threshold sensitivity, causal CTCA behavior, SAERF descriptor analysis, and model-side versus end-to-end latency boundaries. The reproducibility package is organized to provide code, configuration files, split identifiers, evaluation scripts, diagnostic-statistic scripts, and illustrative sample annotations; redistribution of the full curated benchmark is handled through institutional data-review approval or controlled access when direct video release is restricted. Full article

Detecting unmanned aerial vehicles (UAVs) remains a difficult task, primarily due to their tiny size, rapid motion, and complex backgrounds. Fusing visible and infrared imagery offers complementary advantages for robust detection, yet existing methods rely on spatial feature aggregation that overlooks spectral disparities, coupling noise with textures. Moreover, the small scale and high dynamics of UAVs hinder standard convolution from decoupling target signals from background interference due to limited receptive fields. To solve these limitations, the Wavelet-guided Frequency–Spatial Decoupling Network (WFSD-Net) is designed for visible–infrared UAV detection. First, to tackle fusion noise, the Discrete Wavelet Band-Differentiated Fusion (DWBF) module is designed to explicitly decouple noise-dominant sub-bands from information-rich components by performing spectral decomposition. It aligns low-frequency distributions via adaptive spatial weighting and disentangles high-frequency details using physics-aware rules, achieving source-level noise suppression. Second, an Axial Strip Contextual Attention (ASCA) module is proposed. By utilizing anisotropic strip convolution via orthogonal decomposition, this module captures global contextual dependencies to effectively decouple weak target features from background clutter, enhancing the spatial position encoding capability for weak targets. Finally, the proposed WFSD-Net method is validated on Anti-UAV300 and Multi-Sensor and Multi-View Fixed-Wing UAV (MMFW-UAV) datasets, and experiments demonstrate that the proposed method is superior to existing state-of-the-art (SOTA) methods. Full article

To address the lack of standardized safety assessment tools for upper-limb rehabilitation robots, this study developed an integrated testing platform and a quantitative safety assessment framework, demonstrated with FlexoArm1 as a proof-of-concept. A 6-degree-of-freedom bionic arm equipped with multiple sensors was constructed, and a fuzzy PID control algorithm was employed to improve motion trajectory tracking accuracy. A fuzzy multi-criteria safety assessment model was established by combining the Analytic Hierarchy Process (AHP) with the entropy weight method. Experiments were conducted on the rehabilitation robot FlexoArm1. The platform reliably replaced human subjects in range-of-motion testing, interactive torque measurement (peak torque approximately 6.2 N·m in fully active mode), and spasticity simulation, with angular data showing close agreement with Inertial Measurement Unit (IMU) measurements. The assessment model assigned a comprehensive safety score of 70.23 to the tested device, successfully identifying weaknesses in fault detection capability and structural safety design. The proposed bionic-arm-based testing platform and the accompanying safety assessment methodology provide practical tools and a quantitative basis for standardizing safety evaluation and guiding design optimization of upper-limb rehabilitation robots. Full article

Flexible piezoresistive sensors often face a trade-off between sensitivity and working range. In this work, a gradient porous poly(vinyl alcohol)/carbon black (PVA/CB) composite was fabricated via a simple sugar-templating method. The bilayer structure consists of a small-pore layer and a large-pore layer, enabling sequential deformation under external pressure. As a result, the sensor exhibits a sensitivity of −3.05 kPa⁻¹ in the low-pressure range (0–20 kPa) and maintains a stable response up to 120 kPa. Compared with uniform porous structures, the gradient design shows improved performance in the medium- and high-pressure ranges. The sensor also demonstrates good repeatability, fast response, and stability over 1000 cycles. Practical applications including respiration monitoring, vocal vibration detection, and motion sensing are demonstrated. This work provides a simple and scalable approach for developing flexible pressure sensors. Full article

This paper addresses the challenges of high latency and the lack of an effective recovery strategy in mobile robot target following tasks. In this paper, a low-latency mobile robot target tracking method based on the improved YOLO-World algorithm is proposed. The process primarily consists of three parts: target detection, target tracking, and motion control. First, for target detection, we introduce a tailored lightweight backbone network, GSS, within the YOLO-World framework, which progressively expands the receptive field through cascaded convolutional operations and enhances cross-group feature interaction via a channel mixing mechanism, significantly improving model efficiency with minimal loss in detection accuracy. Additionally, depthwise separable convolution is applied to the detection head to reduce computational redundancy. Secondly, in the target tracking part, a lightweight target tracking algorithm based on improved BoT-SORT is adopted, and the tracking delay is effectively reduced by optimizing the ReID feature extraction backbone network. Then, the motion control part adopts an active search strategy based on visual servo control. When the tracked target is lost, the strategy utilizes a camera motion compensation-based tracker to predict the target motion state and controls the robot to actively search for the target accordingly. Subsequently, feature tracking is resumed through target re-recognition, thus re-establishing target following. Experiments on public datasets and real-world scenarios demonstrate that the proposed method achieves strong robustness and real-time performance. Full article