Search Results (950)

Human detection in drone imagery is an important capability for applications that require information on the location and number of people, including search and rescue, urban pedestrian-flow analysis, crowd-density assessment, and public-space safety management. However, in oblique drone imagery, acquiring and annotating sufficient real data remains difficult because of crowd density, privacy constraints, background clutter, and position-dependent scale variation. This study proposes a copy-and-paste-based synthetic data augmentation procedure for imagery acquired using a DJI Phantom 4 RTK under a 30° camera pitch angle and a 30 m flight altitude, following DJI search-and-rescue observation guidelines. The proposed procedure combines human-free drone background images with a public full-body human dataset and incorporates perspective-based geometry-aware scaling, context-aware placement, and density-matched chip selection based on real person-count distributions. Three YOLO-family detectors (YOLOv8x, YOLO11x, and YOLO26x) and the Transformer-based RT-DETR-X were evaluated at three real-data proportions (10%, 25%, and 100%). Across these twelve combinations, four augmentation strategies (random, geometry-, context-, and combined geometry- and context-aware augmentation) were compared under the same 50% synthetic ratio. For the YOLO family, geometry-aware augmentation improved mAP50-95 in eight of the nine combinations and achieved a mean gain of +0.64 percentage points over the real-only baseline, with the largest gain of +1.99 percentage points under the YOLO11x 25% real-data condition. For RT-DETR-X, the corresponding changes were +0.23, −1.83, and −0.15 percentage points at 10%, 25%, and 100% real data, respectively. These results suggest that, under the acquisition conditions evaluated in this study, aligning pasted person size with the image-position-dependent perspective scale can improve the effectiveness of synthetic augmentation when the chip-level person-count distribution is controlled to match that of the real training data. Full article

Foot-mounted inertial pedestrian navigation is an attractive infrastructure-free solution for pedestrian positioning in environments where Global Navigation Satellite System (GNSS) signals are unavailable, degraded, or intentionally excluded. However, mixed-gait and multi-level motion expose two persistent failure modes: unreliable zero-velocity detection during high-dynamic motion and weak yaw observability after standard zero-velocity-aided filtering. This paper presents a probability-driven adaptive foot-mounted inertial navigation framework that couples a lightweight motion-probability front-end with probability-weighted mode-adaptive zero-velocity detection, local straight-heading constraints, and probability-gated stair-height updates. A 335-dimensional motion-feature LightGBM classifier first outputs six gait probabilities from 128-sample windows of six-axis inertial data. These probabilities continuously determine the zero-velocity detector threshold and the reliability of heading and vertical pseudo-measurements. For planar navigation, a motion-aware local straight-heading constraint (MA-LSHC) establishes local reference headings only over curvature-safe straight segments, and an adaptive heading covariance controls the strength of each heading pseudo-measurement. For stair navigation, height changes are quantized into integer-riser increments only when stair probabilities and vertical inertial evidence agree. All data were collected in outdoor pedestrian environments, and the evaluation includes four parts: motion classification, training and detector diagnostics, closed rectangular planar navigation, and multi-level stair-height reconstruction. The motion-feature classifier achieves 96.24% accuracy and 96.17% macro-F1. In a mixed-gait closed rectangular route, the proposed method obtains a 1.09 m closure error and a 0.68% relative closure error, outperforming fixed SHOE-ESKF, Wang-AWGF, Guo-SIHDC, and Deng-HDR baselines. In stair experiments, the proposed method achieves a grid-node RMSE of 0.13 m and correctly identifies 38 ascending and 38 descending risers. These results demonstrate that motion probabilities can be used not merely as labels, but as continuous reliability cues for adaptive inertial constraints. Full article

Pedestrian-future-trajectory prediction is a critical task in intelligent driving, traffic-scene understanding, and active safety decision-making. In real-world road environments, pedestrians’ historical trajectories are frequently affected by occlusions from vehicles, other pedestrians, road infrastructure, and detector-missed detections, leading to missing values in the historical-observation sequence. Understanding and handling missing values in pedestrian-observation sequences is essential for improving the performance of prediction models; however, existing research has rarely considered this realistic scenario. This paper proposes RCN-MAMBA, a pedestrian-future-trajectory-prediction method for occlusion scenarios. The overall pipeline follows a de-occlusion completion first, future-trajectory-prediction second paradigm. Specifically, in the de-occlusion stage, based on linear interpolation for trajectory completion, a BiLSTM Residual-Correction Network is employed to further refine the occluded trajectory. This paper further proposes the C_MAMBA prediction architecture: first, explicit motion-state encoding is introduced to inject position, velocity, and acceleration information explicitly into the trajectory representation; then, multi-scale temporal convolution is used to extract local motion patterns across different time ranges; finally, a Temporal Bi-Mamba module is introduced to model long-term temporal dependencies in the completed trajectory from both forward and backward temporal directions. RCN-MAMBA achieves competitive performance on the JAAD dataset. Qualitative results show that its predicted future trajectories are consistent with actual motion trends. Quantitative analysis is conducted under six whole-frame random occlusion ratios: 5%, 10%, 15%, 20%, 30%, and 40%. RCN-MAMBA achieves optimal results under all these occlusion ratios. The paired-t test for each trajectory further confirms that the performance advantage of our proposed method over all baseline methods is statistically significant, demonstrating that the proposed method possesses good prediction accuracy and robustness under varying occlusion intensities. Full article

In environments where global navigation satellite system (GNSS) signals are unavailable or unreliable, inertial measurement unit (IMU)-based pedestrian dead reckoning (PDR) offers considerable potential because it operates independently of external infrastructure. However, maintaining stable and accurate positioning across varying conditions remains challenging. To address this issue, this paper proposes a large language model (LLM)-guided IMU-PDR framework for detection and error correction. The proposed method transforms IMU time-series data into a structured representation tailored to PDR tasks and incorporates domain-specific prior knowledge into prompts to guide the LLM in step-count detection, step-length estimation, heading correction, and subsequent trajectory reconstruction. Experimental results demonstrate that the proposed framework improves step-count detection accuracy, achieves competitive step-length estimation performance, effectively suppresses heading drift, and enhances the overall consistency of reconstructed trajectories. These results suggest that, when guided by domain-specific prior knowledge, LLMs have the potential to interpret and reason effectively over IMU signals, providing a new approach to intelligent detection and error correction in IMU-PDR. Full article

Reliable visual perception is essential for autonomous urban mobility, where dense traffic, severe occlusion, low illumination, and heterogeneous object distributions challenge pedestrian and animal detection. This study proposes an optimization-oriented hybrid visual perception architecture that formulates visual perception as a unified optimization problem by jointly improving object localization, adaptive suppression, confidence calibration, feature refinement, and safety-oriented risk assessment. The proposed architecture integrates YOLOv11 for real-time object detection, Dynamic Non-Maximum Suppression (DNMS) for adaptive overlap filtering, EfficientNetB7 for hierarchical feature refinement, Bayesian confidence fusion for probabilistic confidence recalibration, and a mathematical risk assessment model for real-time decision support. The architecture was evaluated on COCO2017, Open Images Dataset V7, BDD100K, and a harmonized hybrid dataset and compared with representative CNN-based and YOLO-based baselines, including YOLOv5, YOLOv8, MobileNet, ResNet50, DenseNet121, and EfficientNetB7. The experimental results demonstrate that the proposed architecture consistently achieves superior performance, reaching 96% accuracy, 96% F1-score, 81% mAP@0.5:0.95, a 4% Miss Rate, a 3% False Positive Rate, and real-time inference at 25 Frames Per Second (FPS). Furthermore, the mathematical risk assessment model classifies detected objects into Safe, Warning, and Danger categories, enabling interpretable safety-aware decision support. The findings indicate that the proposed architecture offers an effective optimization-based solution for reliable and real-time perception in autonomous urban mobility. Full article

Cooperative perception through vehicle-to-vehicle (V2V) communication can resolve occlusions that single-vehicle systems cannot overcome, yet existing frameworks rely on simulated environments and expensive multi-sensor platforms. This paper presents CoSafe, a cooperative perception and reasoning framework built on real-world data acquired from dashboard cameras with integrated GPS. CoSafe extends a previously validated image-based positioning algorithm by adding YOLOv8 object detection, cooperative state fusion, Chain-of-Thought reasoning using a Large Language Model, and a deterministic rule-based safety validation layer. The core contribution is a spatial–temporal vehicle matching framework that associates frames captured from overlapping geographic locations at different timestamps, enabling cooperative hazard reasoning across asynchronous and partially observable vehicle streams. This paper also introduces the Cooperation Gain metric to quantify the proportion of cases in which hazard detection depends on inter-vehicle information sharing. On a real-world occluded-pedestrian scenario, cooperation provides an advance warning of at least 1.73 s before the following vehicle’s own detector registers the pedestrian, with a Cooperation Gain of 0.533 measured against manual human annotation. An ablation across single-vehicle, rule-only, and LLM-only configurations isolates the contribution of each component, and a negative-scenario test yields zero false alarms in normal traffic. Full article

The Internet of Vehicles (IoV) is revolutionizing intelligent transportation systems by ubiquitous connectivity of vehicles, roadside infrastructure, pedestrians, edge/cloud platforms, and smart-city services. With the IoV evolving towards highly connected, autonomous and data-driven mobility ecosystems, it needs to meet challenging requirements for low latency, scalability, interoperability, security, privacy and trust. This paper presents a comprehensive cross-layer approach for intelligent, secure and privacy-preserving IoV systems. It is built upon an analytical framework and systematically studies the perception, communication, edge/cloud computing, blockchain-enabled trust and application layers of IoV technologies. In addition, the paper presents an in-depth review of the enabling techniques such as machine learning (ML), deep learning (DL), reinforcement learning (RL), federated learning (FL), blockchain, cybersecurity mechanisms, digital twins, edge computing, 6G integration, and resource allocation. Moreover, it discusses the interplay and trade-offs between intelligence, security, privacy, computation, latency, and scalability. The survey also covers other significant challenges like intrusion detection, decentralized authentication, privacy-preserving learning, blockchain overhead, semantic interoperability, post-quantum security, and standardized datasets. This study is intended to serve as a structured reference for the development of scalable, trustworthy, and intelligent IoV systems by highlighting state-of-the-art techniques, open research gaps, and future directions. Full article

Pedestrian-detection research and pre-crash safety assessment predominantly represent upright pedestrians, although prone, supine, lateral, seated, crouched, kneeling, partially collapsed, and fall-transition states alter target geometry, visibility, sensor signatures, and intervention time. Cross-study comparison is further limited by inconsistent posture labels, data provenance, latency boundaries, uncertainty reporting, and vehicle-response assumptions. We developed NUP-REPORT 1.0 as a provisional reporting and benchmarking framework through a structured narrative synthesis of a 45-source derivation corpus covering epidemiology, sensing benchmarks, uncertainty and assurance methods, reporting-guideline methodology, and public safety protocols. A reconstructed decision ledger documented 46 candidate concepts: 30 were retained as checklist items, 10 were assigned to an extended descriptor set, and six were merged. Each retained item was mapped to supporting evidence and classified as universal core (n = 19), component-contingent core (n = 4), or conditional (n = 7). The framework comprises six domains, a scenario-coverage matrix, five non-overlapping event timestamps, detection-referenced stopping equations, and a 30-item checklist. A purposive feasibility audit of 20 publications, including the adjacent pedestrian-detection literature not designed specifically for non-upright evaluation, illustrated checklist use. Within this sample, target orientation and static-versus-transition state were each reported explicitly in five of 20 publications (25%); none of the 17 applicable papers reported both time-to-first-detection and detection distance, none evaluated confidence calibration, and none of the 20 reported independent-unit uncertainty intervals. Vehicle-response items were non-applicable to papers making no intervention claim. These observations are sample-specific and do not estimate field-wide reporting prevalence. NUP-REPORT is not a consensus standard, certification procedure, or safety score; it is a traceable Version 1.0 proposal for study design, retrospective audit, and stakeholder refinement. Full article

Reliable localization and mapping are critical for intelligent robotic systems operating in dynamic indoor environments, where pedestrians and other moving objects can lead to erroneous feature associations, map contamination, and accumulated trajectory drift. To address these challenges, this study proposes the Boundary-Protected Semantic-Geometric Dynamic-Probability (Boundary-SGDP) framework, an enhanced red–green–blue-depth (RGB-D) visual simultaneous localization and mapping (SLAM) system based on boundary-protected semantic–geometric dynamic-probability estimation. The proposed method combines instance-level semantic priors generated by the YOLO26n-seg detector, a segmentation-oriented model in the You Only Look Once (YOLO) family, and the Segment Anything Model 2 (SAM2) with morphological region decomposition and RGB-D depth-edge detection. Potentially dynamic regions are further divided into dynamic interiors, semantic boundary protection bands, and geometrically informative depth-edge regions. Semantic and geometric cues are integrated to estimate a dynamic score for each feature, which is subsequently propagated to the MapPoint level as a dynamic probability. During pose optimization, these probabilities are used to adaptively adjust the weights of reprojection constraints, thereby reducing the influence of motion-contaminated observations while preserving geometrically valuable features around object boundaries and occlusion regions. Unlike conventional hard semantic masking strategies, Boundary-SGDP provides a soft and adaptive mechanism for handling dynamic observations. Experiments conducted on four dynamic walking sequences from the TUM RGB-D benchmark demonstrate that the proposed method achieves lower absolute and relative trajectory errors than the original ORB-SLAM3 system, while retaining substantially more boundary-related features. The results confirm the effectiveness of semantic–geometric fusion and boundary protection for robust visual localization and mapping in dynamic indoor scenes, and demonstrate the potential of the proposed framework for practical autonomous navigation and intelligent perception applications. Full article

Accurate and scalable driver performance assessment is critical for improving road safety and reducing traffic-related injuries and fatalities, particularly in low- and middle-income countries where the majority of global road deaths occur. This paper presents an exploratory proof-of-concept framework for automated driver evaluation that combines real-world dashcam footage, YOLOv8-based object detection, and multimodal large language models (MLLMs), specifically Gemini 1.5 Flash. Two prompting strategies, narrative and rule-based, were designed to assess driver behavior against standardized licensing criteria derived from the California Department of Motor Vehicles (DMV) driving performance evaluation score sheet. The framework was evaluated across 11 manually curated driving scenarios covering intersections, pedestrian crossings, stop signs, cyclists, and emergency vehicles. Ground-truth labels were established through consensus between two traffic engineering experts cross-referencing official California DMV evaluation criteria. In this preliminary evaluation, the rule-based prompt achieved higher agreement with ground-truth assessments (10/11 scenarios, 90.9%) compared to the narrative prompt (7/11 scenarios, 63.6%), particularly in detecting clear rule violations. The narrative approach demonstrated greater contextual flexibility in ambiguous situations. These results should be interpreted as preliminary, given the small sample size, manually curated dataset, and absence of large-scale statistical validation. Nonetheless, the findings illustrate how combining visual detection with structured language-model prompting may support interpretable, policy-aligned driver evaluation. Key limitations include dependence on video quality, limited scenario diversity, absence of temporal behavioral modeling, and reproducibility constraints tied to proprietary API behavior. Future work should expand validation to larger annotated datasets, incorporate temporal sequence modeling, and explore region-specific regulatory adaptation. Full article

Subway tunnels exacerbate fire hazards, revealing a critical lacuna regarding the spatio-temporal coupling of fire progression and human egress. To address this, this study proposes and validates an Integrated Spatio-Temporal Risk Assessment Model to explicitly quantify survivability thresholds under complex fire dynamics. The framework synergizes Large-Eddy Simulation (LES) based Computational Fluid Dynamics with agent-based pedestrian trajectory modeling within a 3D tunnel featuring a 2% longitudinal gradient. Evaluating 9.5 MW and 12 MW fire energies, the model assessed Single-Sided Evacuation (SSE), Double-Sided Evacuation (DSE), and Sprinkler-Assisted Single-Sided Evacuation (SSE-S) across 2160 agents. Hazards were quantified by continuously resolving Fractional Effective Dose (FED) indices, 60 °C boundaries, and 500 ppm CO fronts. Simulations reveal the gradient induces a severe stack effect, accelerating toxic dispersion and yielding temperatures exceeding 1200 °C. Consequently, SSE engendered fatal bottlenecks (FED: 15.33), whereas DSE optimized pedestrian flux, capping peak FED at 0.52. Crucially, while active suppression (SSE-S) extinguished flames within 105 s, thermodynamic cooling induced a paradoxical loss of smoke buoyancy, causing toxic layers to stratify at the breathing zone. Ultimately, while DSE and SSE-S are paramount for survivability, water-based suppression generates localized toxicological risks, necessitating the integration of low-level smoke detection and extraction architectures in future subterranean designs. Full article

Object detection in visible nearshore surveillance imagery is of great importance for maritime safety, intelligent coastal monitoring, and water rescue applications. Nevertheless, reliable detection remains difficult because nearshore scenes often contain numerous small targets, cluttered wave patterns, shoreline textures, and substantial illumination variations. Moreover, existing public datasets mainly emphasize vessel detection and provide limited nearshore object categories. To solve these limitations, this study presents a large-scale visible nearshore dataset containing 20,934 images annotated with seven categories: pedestrian, sailor, swimmer, ship, boat, flotage, and seamark. The dataset is designed to support comprehensive evaluation and fair comparison of detection algorithms in complex nearshore environments. Based on the proposed benchmark, we conduct extensive evaluations of multiple mainstream object detectors and further develop a detection framework termed VN-DETR. The proposed model enhances both feature extraction and multi-scale feature fusion for nearshore scenarios. Specifically, a kernel selective attention based on WTConv (WKSA) module is designed to enlarge the receptive field and exploit contextual information in visible images, enabling more accurate object classification. In addition, a cross-layer feature selection and fusion (CFSF) module is introduced to perform feature matching, selection, and fusion across adjacent layers, enhancing the discriminability between foreground objects and complex nearshore backgrounds. This design effectively improves robustness against background noise such as wave reflections and shoreline textures. Extensive experiments on the constructed dataset demonstrate that VN-DETR consistently outperforms representative baseline methods and achieves superior detection performance, particularly for challenging small object categories. Full article

Proper management of pedestrian flow at urban crosswalks is crucial for alleviating congestion, reducing waiting times, and improving general traffic safety. This paper presents AH-YOLOFlow, a computer vision framework for pedestrian detection, multi-object tracking, and zone-based flow analytics at urban crosswalks. The proposed framework integrates an attention-enhanced YOLO26n detector, Hungarian assignment-based multi-object tracking, polygon-based spatial zone modeling, and key performance indicator aggregation to estimate pedestrian flow rate, waiting behavior, crossing activity, and localized pedestrian density. The framework was evaluated using a 228.74-s CCTV crosswalk video with a fixed split and an augmented training dataset. Experimental results demonstrate that the proposed detector achieved a mAP@50-95 of 0.5794, mAP@50 of 0.8000, precision of 0.9003, and recall of 0.6847, outperforming the evaluated YOLO-family baselines in terms of mAP@50-95 and precision while remaining competitive with RT-DETR. In the end-to-end pedestrian analytics evaluation, AH-YOLOFlow identified 638 zone-entry events and 495 directional movement events, and successfully associated 54.01% of tracked pedestrian identities with predefined spatial regions. Furthermore, a pseudo-MOT evaluation based on spatially propagated identity annotations demonstrated enhanced tracking consistency following data augmentation, achieving a pseudo-MOTA of 0.7671, pseudo-MOTP of 0.9205, pseudo-HOTA of 0.4527, and pseudo-IDF1 of 0.3844. These findings demonstrate that AH-YOLOFlow provides an effective and computationally efficient framework for real-time crosswalk monitoring by integrating pedestrian detection, tracking, and spatial analytics. In future work, we will focus on validating the system across diverse urban environments using larger multi-scene datasets with manually verified identity annotations to further improve tracking robustness and generalization. Full article

Small-scale pedestrians in UAV imagery often exhibit limited pixel coverage, weak texture, and severe background interference, while progressive network downsampling can further degrade their short-side structures and increase missed detections. To address this problem, we propose DGF-YOLO, a degradation-guided feature enhancement method built on YOLOv12n. The method introduces a degradation-level criterion to identify the feature stage at which a pedestrian first undergoes significant structural degradation and, based on the resulting statistics, incorporates a high-resolution P2 detection head. It further employs a Directional Structure-Aware module to enhance local, horizontal, and vertical structural cues through adaptive multi-branch fusion, a Degradation-Guided Attention module to learn a degradation guidance map under explicit supervision and reweight degradation-sensitive regions, and a Fine-Grained Structure Preservation module to retain local contours and contextual details using depthwise and dilated convolutions. On the single-class pedestrian detection task constructed from VisDrone2019-DET, DGF-YOLO achieves 70.4% precision, 51.8% recall, 59.7% mAP50, and 26.9% mAP50-95, improving the YOLOv12n baseline by 6.6, 7.4, 10.3, and 6.5 percentage points, respectively. The results suggest that the proposed feature enhancement strategy helps reduce missed detections associated with structural degradation in small-scale pedestrians. Full article

Pedestrian detection in crowds is a challenging problem in computer vision. Existing occlusion-handling methods heavily rely on expensive visible-box annotations to locate visible body parts, posing severe limitations in label acquisition cost and open-world generalization. To break through this limitation, we propose a novel Structured Topology of Gridpoints (STG) framework. Operating strictly under standard full-box annotations without any extra visibility supervision, STG aims to achieve implicit, fine-grained local semantic compensation. Specifically, we formulate a coarse-to-fine reasoning paradigm consisting of three interactive stages. To mitigate the high spatial complexity and eliminate background redundancy, we first introduce a Saliency-Aware Feature Filtering (SAFF) mechanism, which leverages gridpoint heatmaps to filter out low-confidence pedestrian candidates. Second, a query-guided Across-Instance Feature Interaction (AIFI) model is designed to utilize inter-instance spatial relationships to propagate missing context from highly visible individuals to their occluded neighbors. Finally, we devise a prior-guided Inner-Instance Gridpoints Interaction (I2GI) model to achieve fine-grained structured part-level feature completion, which dynamically aggregates vital localized cues from diverse human parts to reconstruct holistic pedestrian representations. Extensive experiments on the CityPersons, CrowdHuman, and WiderPerson datasets demonstrate the effectiveness and efficiency of our proposed method. Specifically, STG achieves a log-average miss rate of 7.41% on Reasonable and 32.05% on Heavy Occlusion subsets of CityPersons, while running at up to 16 FPS, outperforming existing part-based methods under full-box supervision. Full article