Search Results (122)

Railway track fasteners, serving as critical connecting components, have a reliability that directly impacts railway operational safety. To address the performance bottlenecks of existing detection methods in handling complex scenarios with diverse fastener types and co-occurring multiple defects, this paper proposes a Type-Guided Expert Model-based Fastener Detection and Diagnosis framework (TGEM-FDD) based on You Only Look Once (YOLO) v8. This framework follows a “type-identification-first, defect-diagnosis-second” paradigm, decoupling the complex task: the first stage employs an enhanced YOLOv8s with Deepstar, SPPF-attention, and DySample (YOLOv8s-DSD) detector integrating Deepstar Block, Spatial Pyramid Pooling Fast with Attention (SPPF-Attention), and Dynamic Sample (DySample) modules for precise fastener localization and type identification; the second stage dynamically invokes a specialized multi-label classification “expert model” based on the identified type to achieve accurate diagnosis of multiple defects. This study constructs a multi-label fastener image dataset containing 4800 samples to support model training and validation. Experimental results demonstrate that the proposed YOLOv8s-DSD model achieves a remarkable 98.5% mean average precision at an Intersection over Union threshold of 0.5 (mAP@0.5) in the first-stage task, outperforming the original YOLOv8s baseline and several mainstream detection models. In end-to-end system performance evaluation, the TGEM-FDD framework attains a comprehensive Task mean average precision (Task mAP) of 88.1% and a macro-average F1 score for defect diagnosis of 86.5%, significantly surpassing unified single-model detection and multi-task separate-head methods. This effectively validates the superiority of the proposed approach in tackling fastener type diversity and defect multi-label complexity, offering a viable solution for fine-grained component management in complex industrial scenarios. Full article

Accurate and real-time sperm tracking is essential for automation in Intracytoplasmic Sperm Injection (ICSI) and fertility diagnostics, where maintaining correct identities across frames improves the reliability of sperm selection. However, identity fragmentation, overcounting, and tracking instability remain persistent challenges in crowded and low-contrast microscopy conditions. This study presents a robust two-layer tracking framework that integrates BoT-SORT with an Extended Kalman Filter (EKF) to enhance identity continuity. The EKF models sperm trajectories using a nonlinear state that includes position, velocity, and heading, allowing it to predict motion across occlusions and correct fragmented or duplicate IDs. We evaluated the framework on microscopy videos from the VISEM dataset using standard multi-object tracking (MOT) metrics and trajectory statistics. Compared to BoT-SORT, the proposed EKF-BoT-SORT achieved notable improvements: IDF1 increased from 80.30% to 84.84%, ID switches reduced from 176 to 132, average track duration extended from 74.4 to 91.3 frames, and ID overcount decreased from 68.75% to 37.5%. These results confirm that the EKF layer significantly improves identity preservation without compromising real-time feasibility. The method may offer a practical foundation for integrating computer vision into ICSI workflows and sperm motility analysis systems. Full article

Modern target tracking systems rely on radar as a sensor to detect targets and generate raw track points. These raw track points are affected by the radar’s own noise and the asymmetric non-Gaussian noise resulting from the nonlinear transformation from polar coordinates to Cartesian coordinates. Without effective processing, such data cannot directly support highly reliable situational awareness, early warning decisions, or weapon guidance. Track filtering, as a core component of target tracking, plays an irreplaceable foundational role in achieving real-time, accurate, and stable estimation of moving target states. Traditional deep learning filtering algorithms struggle with capturing long-term dependencies in high-dimensional spaces, often exhibiting high computational complexity, slow response to transient signals, and compromised noise suppression due to their inherent architectural asymmetries. In order to address these issues and balance the model’s high accuracy, strong real-time performance, and robustness, a new trajectory filtering algorithm based on a temporal convolutional network (TCN), Residual Gated Recurrent Unit (ResGRU), and multi-head attention (MHA) is proposed. The TCN-ResGRU-MHA hybrid structure we propose combines the parallel processing advantages and detail-capturing ability of a TCN with the residual learning capability of a ResGRU, and introduces the MHA mechanism to achieve adaptive weighting of high-dimensional features. Using the root mean square error (RMSE) and Euclidean distance to evaluate the model effect, the experimental results show that the RMSE of TCN-ResGRU-MHA is 27.4621 (m) lower than CNN-GRU, which is an improvement of 15.99% in the complex scene of high latitude, and the distance is 37.906 (m) lower than CNN-GRU, which is an improvement of 18.65%. These results demonstrate its effectiveness in filtering and tracking tasks in high-latitude complex scenarios. Full article

Pedestrian trajectory prediction from a vehicle-mounted perspective is essential for autonomous driving in complex urban environments yet remains challenging due to ego-motion jitter, frequent occlusions, and scene variability. Existing approaches, largely developed for static surveillance views, struggle to cope with continuously shifting viewpoints. To address these issues, we propose V-PTP-IC, an end-to-end framework that stabilizes motion, models inter-agent interactions, and fuses multi-modal cues for trajectory prediction. The system integrates Simple Online and Realtime Tracking (SORT)-based tracklet augmentation, Scale-Invariant Feature Transform (SIFT)-assisted ego-motion compensation, graph-based interaction reasoning, and multi-head attention fusion, followed by Long Short-Term Memory (LSTM) decoding. Experiments on the JAAD and PIE datasets demonstrate that V-PTP-IC substantially outperforms existing baselines, reducing ADE by 27.23% and 25.73% and FDE by 33.88% and 32.85%, respectively. This advances dynamic scene understanding for safer autonomous systems. Full article

In the research field of multi-robot cooperation, reliable and low-cost motion capture is crucial for system development and validation. To address the high costs of traditional motion capture systems, this study proposes a real-time 6D pose estimation and tracking method for multi-robot systems based on YolPnP-FT. Using only an Intel RealSense D435i depth camera, the system achieves simultaneous robot classification, 6D pose estimation, and multi-target tracking in real-world environments. The YolPnP-FT pipeline introduces a keypoint confidence filtering strategy (PnP-FT) at the output of the YOLOv8 detection head and employs Gaussian-penalized Soft-NMS to enhance robustness under partial occlusion. Based on these detection results, a linearly weighted combination of Mahalanobis distance and cosine distance enables stable ID assignment in visually similar multi-robot scenarios. Experimental results show that, at a camera height below 2.5 m, the system achieves an average position error of less than 0.009 m and an average angular error of less than 4.2°, with a stable tracking frame rate of 19.8 FPS at 1920 × 1080 resolution. Furthermore, the perception outputs are validated in a CoppeliaSim-based simulation environment, confirming their utility for downstream coordination tasks. These results demonstrate that the proposed method provides a low-cost, real-time, and deployable perception solution for multi-robot systems. Full article

Accurate short-term load forecasting (STLF) is essential for energy management in buildings, yet remains challenging due to the nonlinear interactions among weather, occupancy, and operational patterns. This study presents a reproducible forecasting pipeline applied as a case study to a single anonymized hotel in Greece, representing a highly variable building-scale load. Three heterogeneous data streams were programmatically ingested and aligned: distribution-operator smart meter telemetry (15 min intervals aggregated to daily active energy), enterprise guest-night counts as an occupancy proxy, and meteorological observations from the National Observatory of Athens (NOA). Following rigorous preprocessing, feature construction incorporated lagged demand, calendar encodings, and exogenous drivers. Forecasting was performed with a stacked LSTM architecture (BiLSTM → LSTM → LSTM with dropout and a compact dense head), trained and validated under a leakage-safe chronological split. A bounded random hyperparameter search of forty configurations was tracked in MLflow 3.5.0 to ensure full reproducibility. The best model achieved RMSE of 4.71 kWh, MAE of 3.48 kWh, and MAPE of 3.29% on the hold-out test set, with stable training and robust diagnostics. The findings confirm that compact recurrent networks can deliver accurate and transparent hotel-level forecasts, providing a practical template for operational energy planning and sustainability reporting. Future research should extend this case study to multi-building portfolios and hybrid deep learning architectures. Full article

Trajectory tracking is a key technology for electrical-driven tracked unmanned vehicles (TUVs), while the control model has a significant impact on tracking performance. To improve trajectory tracking accuracy for a dual-motor-driven TUV, a data-driven model-based predictive control scheme is proposed in this article. First, a vehicle dynamics model based on the Long Short-Term Memory (LSTM) network is developed for a TUV. The vehicle’s motion states in a subsequent time step are predicted using a sequence of history states and control inputs, while the multi-body dynamics model in the TUV platform are utilized for training and validation. Then, a neural network-based model predictive control (NN-MPC) strategy is designed, employing the trained LSTM model as the prediction model within a receding horizon framework to compute the optimal motor torques for trajectory tracking. Unlike existing learning-based MPC approaches that mainly focus on wheeled vehicles, this work investigates a neural network-enhanced MPC for tracked unmanned vehicles with coupled longitudinal–lateral dynamics. The simulation results demonstrate that, compared to a physics-model based MPC strategy, the proposed NN-MPC reduces the root mean square (RMS) values of lateral error and heading error by 12.1% and 7.9% in a medium-speed scenario and by 80% and 14.0% in a high-speed scenario. The field experiment further verifies the practical feasibility of the proposed control scheme. Full article

Aiming at the problem of model parameter perturbation in vehicle trajectory tracking control, an improved robust model predictive control (RMPC) method is proposed. Based on the two-degree-of-freedom vehicle model and Serret Frenet error model, a multi-cell hypercube vertex modeling is adopted to map the disturbance range of parameters such as vehicle speed and lateral stiffness to a set of vertices, and dynamic linear combination is achieved through normalized weights. The algorithm design mainly focuses on the dual-layer optimization of the switching mechanism, decomposing the infinite time domain problem into finite time domain optimization and terminal constraints. At the same time, it dynamically updates the vertex parameters to match time-varying uncertainties and then combines Lyapunov theory to design a control invariant set. The results show that in complex road conditions and vehicle state transitions, RMPC can reduce the peak lateral deviation from 1.0 m to 0.2 m, converge the heading deviation to within 2 deg, and significantly reduce the mean and root mean square values of control errors compared to traditional MPC, under the influence of vehicle model parameter perturbations. RMPC has good robustness and real-time performance. Full article

Automated safety monitoring on construction sites requires precise helmet-status detection and robust multi-object tracking in long, occlusion-rich video sequences. This study proposes a two-stage framework: (i) a YOLOv5 model enhanced with self-adaptive coordinate attention (SACA), which incorporates coordinate-aware contextual information and reweights spatial–channel responses to emphasize head-region cues—SACA modules are integrated into the backbone to improve small-object discrimination while maintaining computational efficiency; and (ii) a DeepSORT tracker equipped with fuzzy-logic gating and temporally consistent update rules that fuse short-term historical information to stabilize trajectories and suppress identity fragmentation. On challenging real-world video footage, the proposed detector achieved a mAP@0.5 of 0.940, surpassing YOLOv8 (0.919) and YOLOv9 (0.924). The tracker attained a MOTA of 90.5% and an IDF1 of 84.2%, with only five identity switches, outperforming YOLOv8 + StrongSORT (85.2%, 80.3%, 12) and YOLOv9 + BoT-SORT (88.1%, 83.0%, 10). Ablation experiments attribute the detection gains primarily to SACA and demonstrate that the temporal consistency rules effectively bridge short-term dropouts, reducing missed detections and identity fragmentation under severe occlusion, varied illumination, and camera motion. The proposed system thus provides accurate, low-switch helmet monitoring suitable for real-time deployment in complex construction environments. Full article

The air-conditioner noise in a rehabilitation room can seriously affect the mental state of patients. However, the existing single-layer active noise control (ANC) filters may fail to attenuate the complicated harmonic noise, and the deep recursive ANC method may fail to work in real time. To solve the problem, in a bending-pipe model, a new hybrid adaptive self-loading filtered-x least-mean-square (ASL-FxLMS) and convolutional neural network-gate recurrent unit (CNN-GRU) network is proposed. At first, based on the recursive GRU translation core, an improved CNN-GRU network with multi-head attention layers is proposed. Especially for complicated harmonic noises with more or fewer frequencies than harmonic models, the attenuation performance will be improved. In addition, its structure is optimized to decrease the computing load. In addition, an improved time-delay estimator is applied to improve the real-time ANC performance of CNN-GRU. Meanwhile, an adaptive self-loading FxLMS algorithm has been developed to deal with the uncertain components of complicated harmonic noise. Moreover, to achieve balance attenuation, robustness, and tracking performance, the ASL-FxLMS and CNN-GRU are connected by a convex combination structure. Furthermore, theoretical analysis and simulations are also conducted to show the effectiveness of the proposed method. Full article

Natural latex harvest is pushing the boundaries of unmanned agricultural production in rubber milk collection via integrated robots in hilly and mountainous regions, such as the fixed and mobile tapping robots widely deployed in forests. As there are bad working conditions and complex natural environments surrounding rubber trees, the real-time and precision assessment of rubber milk yield status has emerged as a key requirement for improving the efficiency and autonomous management of these kinds of large-scale automatic tapping robots. However, traditional manual rubber milk yield status detection methods are limited in their ability to operate effectively under conditions involving complex terrain, dense forest backgrounds, irregular surface geometries of rubber milk, and the frequent occlusion of rubber milk bowls (RMBs) by vegetation. To address this issue, this study presents an unmanned aerial vehicle (UAV) imagery rubber milk yield state detection method, termed YOLOv8n-RMB, in unstructured field environments instead of manual watching. The proposed method improved the original YOLOv8n by integrating structural enhancements across the backbone, neck, and head components of the network. First, a receptive field attention convolution (RFACONV) module is embedded within the backbone to improve the model’s ability to extract target-relevant features in visually complex environments. Second, within the neck structure, a bidirectional feature pyramid network (BiFPN) is applied to strengthen the fusion of features across multiple spatial scales. Third, in the head, a content-aware dynamic upsampling module of DySample is adopted to enhance the reconstruction of spatial details and the preservation of object boundaries. Finally, the detection framework is integrated with the BoT-SORT tracking algorithm to achieve continuous multi-object association and dynamic state monitoring based on the filling status of RMBs. Experimental evaluation shows that the proposed YOLOv8n-RMB model achieves an AP@0.5 of 94.9%, an AP@0.5:0.95 of 89.7%, a precision of 91.3%, and a recall of 91.9%. Moreover, the performance improves by 2.7%, 2.9%, 3.9%, and 9.7%, compared with the original YOLOv8n. Plus, the total number of parameters is kept within 3.0 million, and the computational cost is limited to 8.3 GFLOPs. This model meets the requirements of yield assessment tasks by conducting computations in resource-limited environments for both fixed and mobile tapping robots in rubber plantations. Full article

To enhance the path-tracking accuracy and control stability of articulated underground vehicles navigating high-curvature tunnels, this paper proposes a novel Multi-Time-Scale Nonlinear Model Predictive Control (MTS-NMPC) strategy. The core innovation lies in its dynamic adaptation of the prediction horizon to simultaneously compensate for the body torsion effects and yaw deviations induced by high-speed cornering. A high-fidelity vehicle dynamics model is first established. Subsequently, an adaptive mechanism is designed to adjust the prediction horizon based on the reference speed and road curvature. Experimental results demonstrate that the proposed MTS-NMPC achieves remarkable reductions of 35% and 17% in the maximum lateral tracking error and heading deviation, respectively, compared to conventional NMPC. It also improves stability by suppressing the velocity fluctuations of the articulated joint. The superior control performance and robustness of our method are further validated through field tests in an underground mine. Full article

Laser direct energy deposition (LDED) has been widely employed in surface modification and remanufacturing. Achieving high-precision geometries and superior mechanical properties in cladding layers remains a persistent research focus. In this study, an Fe-based alloy was deposited on an AISI 1045 substrate via a wide-beam laser cladding system. Single-track multi-layer samples were prepared with varying z-increment (Z_d), interlayer dwell time (T_I) and laser scanning speed (V) values. The geometry, microstructure, microhardness and wear resistance of the samples were analyzed. Experimental results showed that an estimated Z_d can ensure a constant standoff distance of the laser head and resulting geometric accuracy improvement. Planar grains form at the layer–substrate bonding interface and transition to columnar grains adjacently, while dendrites and equiaxed grains are distributed in the middle and top regions of the layer. The coating layer exhibits much better wear resistance and friction properties than the substrate. The cooling rate can be substantially increased by either raising V or prolonging T_I, resulting in refined grain structures and enhanced microhardness. Real-time monitoring and controlling the mean cooling rate have been demonstrated to be effective strategies for enhancing cladding layer performance. Full article

Aiming at the problems of excessive computational load, insufficient real-time performance, and an excessive amount of model parameters in track inspection, this paper proposes a lightweight track feature detection module (YOLO-LWTD) based on YOLO11n: first, the StarNet module is integrated into the backbone network, and its elemental multiplication operation is utilized to enhance the feature characterization capability; second, in the neck part, a lightweight extended path aggregation network reconstructs the feature pyramid information flow paths by combining with the C3K2-Light module to enhance the efficiency of the multi-scale feature fusion; finally, in the head part, a lighter and more efficient detection header, Detect-LADH, is used to reduce the feature decoding complexity. Experimental validation showed that the improved model outperforms the benchmark model in precision, recall, and mean average precision (MAP) by 0.5%, 2.0%, and 0.8%, respectively, with an inference speed of 163 FPS (a 38.1% improvement). The model volume is compressed to 1.5 MB (a 71.1% lightweight rate). This provides an energy-efficient solution for lightweight track detection tasks geared towards embedded deployment or real-time processing. Full article

Precise alignment between virtual reality (VR) and haptic interfaces is essential for delivering an immersive visuo-haptic mixed reality experience. Existing methods typically depend on markers, external trackers, or cameras, which can be intrusive and hinder usability. In addition, previous network-based approaches generally rely on image data for alignment. This paper introduces a deep neural network-based alignment method that eliminates the need for such external components. Unlike existing methods, our approach is designed based on coordinate transformation and leverages a network model for alignment. The proposed method utilizes the head-mounted display (HMD) position, fingertip position obtained via hand tracking, and the six-degrees-of-freedom (6-DOF) pose of a haptic device’s end-effector as inputs to a neural network model. A shared multi-layer perceptron and max pooling layer are employed to extract global feature vectors from the inputs, ensuring permutation invariance. The extracted feature vectors are then processed through fully connected layers to estimate the pose of the haptic device’s base. Experimental results show a mean positional error of 2.718 mm and the mean rotation error of 0.5330°, which equates to 1.3% relative to the haptic device’s maximum length. The proposed method demonstrates robustness against noise, demonstrating its applicability across various domains, including medical simulations, virtual prototyping, and interactive training environments. Full article