Search Results (2,253)

Visual–Inertial Odometry (VIO) has become a key technology for motion estimation in robotics and autonomous systems. However, deploying accurate VIO pipelines on embedded platforms remains challenging due to the trade-off between estimation accuracy, real-time performance, and energy consumption. This paper presents a hardware-aware Feature-PLPD VIO framework that integrates a point-and-line visual front-end with a loosely coupled Error-State Extended Kalman Filter (ESEKF) back-end. The proposed approach follows Algorithm Architecture Adequacy (A3) principles to preserve estimation robustness while limiting computational and memory requirements. To investigate its scalability across the considered resource constraints, two end-to-end embedded implementations are developed: a performance-oriented stereo VIO system on a GPU-based platform using GPU-aware software design, and a frugality-oriented low-cost RGB-D-assisted monocular VIO system on an FPGA-based architecture using hardware–software co-design with depth scale correction. The stereo GPU-based implementation is evaluated offline in both outdoor and indoor environments and is additionally validated through real-time on-the-fly deployment on a Scout Mini robot, whereas the FPGA-based implementation is evaluated offline using the indoor VICON dataset. Experimental results demonstrate meter-level trajectory accuracy and real-time performance under strict resource constraints. Averaged over four KITTI sequences, the proposed ESEKF-based fusion reduces the translation and rotation ATE by approximately 25% and 18%, respectively, compared with the corresponding VO-only configuration, while a 10% reduction in translation ATE is achieved in the indoor VICON environment, resulting in a normalized ATE of $5.09\%$ over the $23.77$ m trajectory. Both embedded implementations sustain around 20 fps, and the architectural evaluation highlights complementary accuracy–runtime–energy trade-offs between the GPU- and FPGA-based solutions. These results demonstrate the feasibility of scaling the proposed embedded VIO framework toward resource- and energy-constrained robotic applications under the investigated experimental and hardware configurations. Full article

Upper limb spasticity is a common complication of stroke that negatively affects motor function, range of motion, and quality of life. Although abobotulinumtoxinA is widely used for the treatment of focal post-stroke spasticity, evidence regarding its early onset of action remains limited. This prospective observational pilot study investigated the early clinical, functional, and neurophysiological effects of abobotulinumtoxinA in 20 adults with post-stroke upper limb spasticity. Patients received ultrasound-guided injections into the biceps brachii and flexor digitorum superficialis muscles and were evaluated at baseline, 24 h, 5 days, 7 days, 14 days, and 30 days after treatment. Outcome measures included the Modified Ashworth Scale (MAS), passive range of motion (ROM), caregiver-reported hand function assessed using a visual analog scale (VAS), and compound muscle action potential (cMAP) recordings. No significant changes were observed at 24 h. A significant reduction in spasticity was detected at the day-5 assessment, accompanied by significant improvements in ROM and functional outcomes. Neurophysiological assessment demonstrated significant reductions in cMAP amplitudes at the day-5 assessment, indicating early neuromuscular transmission blockade. Clinical improvements were also first detected at day-5 assessment and were maintained at subsequent evaluations through day 30. These findings indicate that clinically meaningful effects of abobotulinumtoxinA were detectable by the first scheduled assessment at day 5 after injection. Given the absence of intermediate assessments between 24 h and day 5, the exact timing of treatment onset cannot be determined. A multimodal assessment approach may help characterize early treatment-related changes and provide useful information for future studies investigating the timing of adjunctive rehabilitation interventions. Full article

The integration of AI vision-based sensing and human motion analysis has grown to be a key element of intelligent perception systems, which now allow for automated interpretation of human movement, activity patterns, and complex visual behaviors. However, accurate functional movement assessment from monocular aerial and ground-view videos remains challenging due to low spatial resolution, background clutter, occlusions, and large variations in body posture, limiting the reliability of AI-assisted future healthcare applications. This study presents a multi-level framework that integrates asymmetric deep feature representation with transformer-based architecture and graph-driven optimization for robust vision-based human movement analysis. First, a Heavy Attention Transformer is employed to enhance image quality and emphasize clinically relevant anatomical and motion patterns by suppressing background interference. Panoptic segmentation and Real-Time Detection Transformer V2 are then used for subject localization, followed by skeletal keypoint extraction using YOLOv8. The proposed framework adopts an asymmetric dual-stream feature extraction strategy, where global contextual information is captured through Bag of Visual Words, Video Swin Transformer, Video Masked Autoencoder, and TimeSformer, while local biomechanical motion dynamics are modeled using DiffPose, PoseFormer, and Spatial–Temporal Graph Convolutional Networks. The key contribution lies in the asymmetric feature design that preserves the distinct information structures of visual context and skeletal dynamics. To reduce feature redundancy and select discriminative clinical representations, the Slime Mould Algorithm is utilized as a metaheuristic optimizer. The optimized features are subsequently classified using a Graph Attention Network for automated functional movement assessment. Experimental evaluation on the UAV-Human and UCF-ARG benchmark datasets achieved an accuracy of 82.50% and 78.20%, respectively. The proposed framework illustrates the potential of asymmetry-aware AI-enabled vision sensing to perform strong human movement analysis in complex viewpoints and lays the groundwork for future healthcare-related applications such as remote human movement evaluation and rehabilitation monitoring. Full article

Gas-insulated substations (GISs) have been widely adopted in modern power systems due to their compact design and high reliability. However, the potential generation of toxic byproducts poses significant risks to manual operations such as gas pressure adjustments, highlighting the necessity for robotic deployments. This paper proposes a Robot Operating System (ROS)-based robotic framework with human–robot collaborative autonomy for GIS operations. The framework employs a 6-degree-of-freedom (6-DOF) robotic arm, integrated with a motion control system for trajectory execution, a visual perception system enhanced by the coordinate attention (CA) mechanism for small-scale detection and localization, and a communication system for data exchange. The framework improves the accuracy of component perception and enables fine manipulation in complex environments, reducing reliance on manual intervention while facilitating a safe collaborative autonomous workflow through dynamic adjustment of autonomy level and control authority. Experimental results on the representative gas pressure adjustment task demonstrate an autonomous operational success rate exceeding 90% under the proposed configuration in dynamic scenarios. By enhancing safety and precision, this study advances robotic solutions for hazardous operations and lays a foundation for broader infrastructure applications. Full article

Water-surface object detection remains challenging under strong reflections, wave disturbances, adverse weather, low illumination, and distant or small targets, which can severely degrade or obscure discriminative visual cues. Although radar–camera fusion provides complementary geometric and motion information, sparse and uncertain radar observations may introduce unreliable cues, while camera–radar temporal asynchrony can further reduce cross-modal spatial consistency. To address these issues, this paper proposes Radar Prior and Time-Lag-Aware Quality-Adaptive Fusion (RPT-Fusion), a radar–camera fusion framework that preserves the camera as the primary semantic modality while treating radar as a reliability-controlled auxiliary source. Sparse 4D radar measurements are transformed into probabilistic occupancy, density, velocity, and reliability priors, with direction-dependent spatial uncertainty represented through anisotropic radar-prior modeling. Local, global, and temporal quality estimates are then jointly used to regulate radar contributions during multi-scale feature fusion. In particular, the measured camera–radar time lag is explicitly incorporated into both radar-prior construction and feature-level reliability control. Experiments on WaterScenes show that RPT-Fusion achieves 92.31% mAP50 at an Intersection-over-Union (IoU) threshold of 0.50 and 68.23% mAP50–95 averaged over IoU thresholds from 0.50 to 0.95, outperforming WS-DETR by 0.82 and 3.69 percentage points, respectively. Ablation experiments verify the contributions of radar-prior modeling, quality-adaptive fusion, and temporal quality, while repeated-training experiments show low performance variability. Time-lag analysis reveals condition-dependent benefits, with the largest observed improvement of 1.15 percentage points occurring when the camera–radar offset reaches or exceeds 20 ms. Radar-frame-dropping experiments further show that mAP50 decreases from 92.31% to 91.54% when radar observations are progressively removed, indicating that the camera-centric pathway retains a substantial detection capability under incomplete radar observations. These results demonstrate the effectiveness of reliability-controlled radar assistance for robust radar–camera object detection in complex water-surface environments. Full article

Multi-modal 3D object detection is critical for autonomous driving perception. While Bird’s Eye View (BEV) fusion methods effectively integrate LiDAR and camera features, they primarily focus on single-frame fusion and neglect temporal context. We propose CamT-BEV, a camera-temporal-enhanced BEV fusion framework for improved multi-modal 3D object detection. Our key insight is that temporal modeling is particularly critical for the camera branch to resolve monocular depth ambiguity and object occlusion, while single-frame LiDAR representation already provides accurate instantaneous geometry. We thus propose a camera-centric temporal enhancement module via ego-motion warping and ConvLSTM temporal encoding. Extensive experiments on the nuScenes dataset demonstrate that CamT-BEV achieves competitive perception performance, attaining 0.6971 NDS and 0.6683 mAP, with notable relative AP gains on challenging categories such as bicycles (+27.3%) and motorcycles (+7.66%) evaluated under category-level mAP (averaged across 0.5 m to 4.0 m distance thresholds). Furthermore, evaluations under fog and miss-beam conditions in nuScenes-C confirm its improved robustness against specific visual and sensor degradations. Crucially, these gains are achieved with low additional computational and memory overhead, demonstrating that targeted camera-temporal fusion is a practical solution for 3D perception. Full article

Visual–inertial odometry (VIO) on industrial stereo camera–IMU platforms suffers from frame-dependent timing errors and degraded feature tracking under auto-exposure, auto-gain adjustment, and rapid motion. This paper presents Hardware-Synchronized Exposure-State-Aware VINS-Fusion (HS-ES-VINS), a hardware-synchronized, exposure-aware extension of VINS-Fusion. The effective sampling time of each global-shutter image is reconstructed at the exposure midpoint using synchronized trigger timestamps, the camera debounce delay, and the frame-wise exposure duration. An exposure–gain–motion risk model adaptively adjusts feature detection criteria, Kanade–Lucas–Tomasi (KLT) tracking parameters, and correspondence filtering, while IMU-integrated rotation supplies initial predictions for feature tracking. On hardware-synchronized indoor and outdoor datasets with independent ground truth, controlled experiments show that the exposure-midpoint timestamp removes the frame-dependent exposure component of the camera–IMU offset that neither fixed nor online scalar compensation can remove, reducing the ATE RMSE by 28.6% (fixed offset) and 22.7% (online estimation) relative to the exposure-start timestamp under 20 ms low-light exposure. Compared with baseline VINS-Fusion, the complete adaptive front end reduces the ATE RMSE by 16.8% and 12.9% on indoor multi-floor and outdoor cycling sequences, and the risk score explains frame-wise tracking quality (correlation of 0.587 with the inlier ratio). The front end runs faster than the baseline (17.1 ms per frame), confirming real-time operation and improved robustness against variable illumination and rapid motion. Full article

Aiming at the application requirements of artificial intelligence and robotics, this paper adopts a lightweight and low-cost scheme to construct a dancing-robot system integrating machine vision, mechanical structure and motion control so as to improve its accurate human-motion imitation capability. It provides an important reference for the engineering application of motion imitation of humanoid robots. A human-pose visual-recognition model is built based on Python and deep-learning techniques. A monocular camera captures two-dimensional images, from which skeletal key-point coordinates are extracted via 3D reconstruction. Joint angles are calculated by inverse kinematics to supply core input data for motion imitation. Mechanically, a humanoid joint structure composed of 16 servos is adopted to realize one-to-one mapping and execution of joint angles. In hardware, a collaborative control circuit is constructed with the main-control module and servo-drive module as the core. Software implements data-parsing, instruction generation and closed-loop control. After system integration and debugging, the robot can stably and accurately reproduce simple human dance movements, which verifies the feasibility and effectiveness of the proposed scheme. Full article

The persistent challenge in visual object tracking, particularly for small and fast-moving targets, lies in the trade-off between effective resolution and contextual information. Fixed search regions cannot adapt to variations in target scale and motion, often resulting in degraded target representation and tracking failures. In this paper, we introduce AdaSAM2, a framework that formulates adaptive search-region selection as a sequential decision-making problem. Unlike conventional per-frame heuristics, our approach employs an event-driven reinforcement learning policy that selects the cropping scale only during initialization and unreliable tracking states, while reusing the previous configuration during stable tracking. To handle target disappearance, we further introduce a lost-aware recovery mechanism that combines progressive search-region enlargement with constrained policy re-selection. Extensive experiments on TSFMO, LaTOT, UAV123, and UAVDT demonstrate consistent improvements over the SAMITE baseline. For example, AdaSAM2 improves Success and Precision on TSFMO from 42.6% and 74.0% to 43.9% and 75.3%, respectively, while improving UAV123 Success and Precision from 69.6% and 92.7% to 71.1% and 94.8%. Moreover, the RL policy is activated on only 0.74% of processed frames on TSFMO, resulting in an average overhead of only 0.0085 ms per frame. These results demonstrate that adaptive input-space optimization can improve tracking accuracy while introducing negligible computational overhead. Full article

Background: Total knee arthroplasty (TKA) is an established treatment for end-stage rheumatoid arthritis (RA). Although cemented fixation has traditionally been preferred because of concerns regarding poor bone quality, advances in implant design and osseointegration have renewed interest in cementless fixation. This study evaluated the long-term clinical and radiological outcomes of cementless TKA in patients with RA. Methods: A retrospective single-center study included adult patients with RA who underwent primary cementless TKA between 2004 and 2021 using the same cruciate retaining implant. Clinical outcomes were assessed using range of motion (ROM), Visual Analog Scale (VAS), Oxford Knee Score (OKS), and patient satisfaction. Radiographic evaluation assessed implant fixation, while implant survivorship was analyzed using bilateral clustering via a Marginal Cox model analysis. Results: Seventy-five cementless TKAs performed in 52 patients were analyzed after a mean clinical follow-up of 10.8 ± 3.8 years. Mean ROM was 101° ± 26.5°, mean VAS score was 0.88 ± 1.84, and mean OKS was 40.3 ± 7.1. Overall, patients were satisfied or very satisfied in 89.6% of implants. Radiographic assessment showed no evidence of progressive radiolucent lines, osteolysis, implant subsidence, aseptic loosening, or malalignment. Only one revision was required because of periprosthetic joint infection, resulting in an implant survivorship of 98.3% at long-term follow-up. Conclusions: Cementless TKA demonstrated excellent long-term implant survival, durable radiographic fixation, and favorable clinical and functional outcomes in patients with rheumatoid arthritis. These findings support modern cementless fixation as a reliable and effective alternative to cemented TKA in carefully selected RA patients. Full article

While soft hand exoskeleton robots have approached human-level dexterity in terms of degrees of freedom, precise control methods for fine motor movements remain a significant challenge. Surface electromyography (sEMG) is widely employed in gesture recognition to enable patients to independently control a soft hand exoskeleton. However, individual finger control remains challenging through sEMG-based control due to the complexity of decoupling synergistic muscle activities. In this study, we propose a muscle fiber-based ultrasound perception strategy for fine hand motion recognition and soft hand exoskeleton control. Ultrasound imaging enables non-invasive visualization of forearm muscle morphology and provides information associated with underlying muscle-fiber activity. By reconstructing muscle morphology from ultrasound images, biologically relevant muscle-fiber features are extracted and fused to characterize fine hand movements. A lightweight Random Forest classifier is subsequently employed to map these biologically informed features to discrete hand actions, providing a computationally efficient recognition module for real-time control. To the best of our knowledge, publicly available ultrasound image datasets specifically designed for fine hand gesture recognition in rehabilitation applications remain limited. In the experiments, a dataset containing 21 hand gestures based on muscle ultrasound images was constructed to evaluate the proposed method. All data were collected from healthy participants as a preliminary proof-of-concept investigation. The results show that the proposed approach achieves an average recognition accuracy of 95.24% across three subjects in finger motion recognition. This preliminary study demonstrates the potential of machine learning-based ultrasound perception for improving fine hand gesture recognition and providing an intuitive control interface for soft hand exoskeletons, thereby enhancing their applicability in hand rehabilitation scenarios. Full article

Textile-based sensors offer a promising platform for human-motion monitoring because of their flexibility, comfort, and ease of integration into clothing. The objective of this study was to develop a textile-based hybrid piezoresistive sensor by combining carbon black (CB) and carbon nanofibers (CNFs) with different filler geometries within a flexible poly[styrene-b-(ethylene-co-butylene)-b-styrene] (SEBS) matrix. The novelty of the proposed approach lies in incorporating CB and CNFs into the same elastomeric sensing layer directly deposited onto an elastic knitted textile substrate and systematically comparing its sensing behavior with the corresponding single-filler CB/SEBS and CNF/SEBS systems. Three sensing coatings, CB/SEBS, CNF/SEBS, and hybrid CB+CNF/SEBS, were deposited onto textile substrates by blade coating. All formulations exhibited uniform coating and good adhesion to the fabric. Coating uniformity and adhesion were assessed through visual inspection, electron microscopy, and mechanical cycling by monitoring peeling, cracking, and delamination. Dynamic piezoresistive tests demonstrated that all sensors were capable of strain monitoring, while the 2 wt% CB+ 2 wt% CNF hybrid formulation exhibited the highest sensitivity among the investigated systems. The resistance response was dependent on both the magnitude and rate of strain deformation. The hybrid sensor was further successfully applied to monitor knee, wrist, and elbow movements during physical exercise. These findings demonstrate that the combination of CB and CNF fillers within a flexible SEBS matrix provides a promising route toward highly strain-sensitive textile sensors for wearable human-motion monitoring. Full article

Visual ecology explores how animal visual systems are related to light environments and important visual tasks. Anolis is a species-rich genus of small lizards that rely extensively on vision. Different species occupy habitats with different vegetation structure. The spectral properties of the habitats are similar, but total light intensities vary widely. They possess laterally positioned eyes with broad monocular visual fields, and a small region of binocular overlap toward the front. They possess a high-resolution central fovea associated with analysis of important images. There is a second, smaller, temporal fovea located where the visual fields of the two eyes overlap, which is associated with distance perception during prey capture. Color and brightness vision depends on four classes of single cones with different spectral absorbance curves, including one sensitive to the ultraviolet and a set of double cones. Most species studied possess similar sets of cones. Cone responses are modified by oil droplet filters, which are more variable among species and may play some role in adaptations to habitat light. Anoles communicate with a colorful expandable throat fan called the dewlap. Dewlap visibility depends on contrast with the natural background. Red dewlaps are most visible in bright, unshaded habitats. Yellow dewlaps are most visible in darker, shaded habitats. Visual displays include motion patterns of the head, body, and dewlap. The most highly visible movements are rapid up-and-down, start-and-stop patterns. These often occur at the beginning of visual displays and draw the attention of conspecifics to the displaying animal. Most Anolis species share similar visual-system properties. Differences in light environments interact with visual-system responses to influence the physical properties of communication displays and other visually-based behaviors. Full article

Introduction and Clinical Significance: Descemet membrane detachment (DMD) is an uncommon but potentially vision-threatening complication of anterior segment surgery. Although immune-mediated graft rejection is a recognized cause of postoperative graft edema following arcuate keratotomy (AK) in eyes with previous penetrating keratoplasty (PKP), structural complications such as DMD may present with similar clinical findings and require fundamentally different management. We report a case of DMD initially misdiagnosed as acute graft rejection following manual AK in a post-PKP eye. Case Presentation: A 47-year-old female with a 20-year history of PKP in the right eye (RE) presented with decreased vision three weeks after undergoing manual AK using a diamond blade to correct high astigmatism. The procedure was complicated by an intraoperative wound leak requiring suturing. Postoperatively, she developed corneal edema and was initially misdiagnosed with acute graft rejection. Despite treatment with topical and systemic corticosteroids, her condition did not improve. Upon referral to our clinic, slit-lamp examination and anterior segment optical coherence tomography (AS-OCT) revealed a near-total, nonplanar DMD. The patient underwent a single air descemetopexy session involving three sequential intracameral air injections, resulting in Descemet membrane apposition and improvement in graft clarity. At 3-month follow-up, visual acuity had improved from hand motion to 20/200, with IOP within normal limits. Longer-term follow-up was unavailable. Conclusions: DMD should be considered as a cause of graft edema after AK in post keratoplasty eyes especially in the presence of complications. Full article

To address the challenges in multi-modal trajectory prediction for multi-aircraft interactions within non-towered terminal airspace, including the insufficient extraction of long-range temporal dependencies, neglect of physical separation constraints, and barriers to integrating flight intentions and multi-source environmental context, this paper develops a trajectory prediction model integrated with long-range temporal modeling, physics-aware spatial interaction, and intention context enhancement. A parameter-shared ST-Transformer temporal encoder is established to capture long-term motion patterns of aircraft via global multi-head self-attention, and temporal attention pooling is adopted to mitigate error accumulation in long-term prediction. A physical distance-aware ST-GAT module is designed, which embeds the spatial distance prior between aircraft into the attention calculation and leverages distance masks to reduce interference from distant irrelevant aircraft. Furthermore, an intention-aware context enhancement module (IACEM) is proposed. It identifies the distribution of flight phases and adaptively incorporates meteorological information to construct enhanced features embedded with high-level semantics and environmental priors. Finally, a CVAE-based framework is utilized to generate multiple candidate trajectories satisfying kinematic constraints. Multiple verification experiments are carried out on the TrajAir dataset. The experimental results demonstrate that the proposed model outperforms various baseline models in terms of ADE and FDE. Ablation studies and visual analysis verify that the three core modules produce synergistic improvements. The model achieves higher prediction accuracy under scenarios involving 2D complex maneuvers, 3D climbing turns, and dense multi-aircraft interactions, which proves the effectiveness and superiority of the proposed algorithm for trajectory prediction in non-towered terminal airspace. Full article