RF-DET: Refocusing on the small-scale objects using aggregated context for accurate power transmitting components detection on UAV oblique imagery

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2025, Volume and Issue: 220, P. 692 - 711

Published: Jan. 25, 2025

Language: Английский

---

Foreign Object Detection Network for Transmission Lines from Unmanned Aerial Vehicle Images

Drones, Journal Year: 2024, Volume and Issue: 8(8), P. 361 - 361

Published: July 30, 2024

Foreign objects such as balloons and nests often lead to widespread power outages by coming into contact with transmission lines. The manual detection of these is labor-intensive work. Automatic foreign object on lines a crucial task for safety becoming the mainstream method, but lack datasets restriction. In this paper, we propose an advanced model termed YOLOv8 Network Bidirectional Feature Pyramid (YOLOv8_BiFPN) detect Firstly, add weighted cross-scale connection structure head network. bidirectional. It provides interaction between low-level high-level features, allows information spread across feature maps different scales. Secondly, in comparison traditional concatenation shortcut operations, our method integrates scale features through settings. Moreover, created dataset Object Transmission Lines from Drone-view (FOTL_Drone). consists 1495 annotated images six types object. To knowledge, FOTL_Drone stands out most comprehensive field lines, which encompasses wide array geographic diverse Experimental results showcase that YOLOv8_BiFPN achieves average precision 90.2% [email protected] 0.896 various categories objects, surpassing other models.

Language: Английский

---

APF-YOLOV8: Enhancing Multiscale Detection and Intra-Class Variance Handling for UAV-Based Insulator Power Line Inspections

F1000Research, Journal Year: 2025, Volume and Issue: 14, P. 141 - 141

Published: Jan. 28, 2025

Background UAV-based power line inspections offer a safer, more efficient alternative to traditional methods, but insulator detection presents key challenges: multiscale object and intra-class variance. Insulators vary in size due UAV altitude perspective changes, while their visual similarities across types (e.g., glass, porcelain, composite) complicate classification. Methods To address these issues, we introduce APF-YOLO, an enhanced YOLOv8-based model integrating the Adaptive Path Fusion (APF) neck Feature Alignment Module (AFAM). AFAM balances fine-grained detail extraction for small objects with semantic context larger ones through local global pathways by advanced attention mechanisms. This work also introduces Merged Public Insulator Dataset (MPID), comprehensive dataset designed detection, representing diverse real-world conditions such as occlusions, varying scales, environmental challenges. Results Evaluations on MPID demonstrate that APF-YOLO surpasses state-of-the-art models different configurations, achieving at least +2.71% improvement [email protected]:0.9 +1.24% increase recall, maintaining real-time performance server-grade environments. Although adds computational requirements, remain within acceptable limits applications. Future will optimize edge devices techniques pruning lightweight feature extractors, enhancing its adaptability efficiency. Conclusion Combined MPID, establishes strong foundation advancing contributing safer effective monitoring.

Language: Английский

---

Insulator Defect Detection Algorithm Based on Improved YOLOv11n

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1327 - 1327

Published: Feb. 21, 2025

Ensuring the reliability and safety of electrical power systems requires efficient detection defects in high-voltage transmission line insulators, which play a critical role isolation mechanical support. Environmental factors often lead to insulator defects, highlighting need for accurate methods. This paper proposes an enhanced defect approach based on lightweight neural network derived from YOLOv11n architecture. Key innovations include redesigned C3k2 module that incorporates multidimensional dynamic convolutions (ODConv) improved feature extraction, introduction Slimneck reduce model complexity computational cost, application WIoU loss function optimize anchor box handling accelerate convergence. Experimental results demonstrate proposed method outperforms existing models like YOLOv8 YOLOv10 precision, recall, mean average precision (mAP), while maintaining low complexity. provides promising solution real-time, high-accuracy detection, enhancing systems.

Language: Английский

---

Physics-Aware Machine Learning Approach for High-Precision Quadcopter Dynamics Modeling

Drones, Journal Year: 2025, Volume and Issue: 9(3), P. 187 - 187

Published: March 3, 2025

In this paper, we propose a physics-informed neural network controller for quadcopter dynamics modeling. Physics-aware machine learning methods, such as networks, consider the UAV model, solving system of ordinary differential equations entirely, unlike proportional–integral–derivative controllers. The more accurate control action on reduces flight time and power consumption. We applied our fractional optimization algorithms to decreasing solution error dynamics. Including advanced optimizers in reinforcement achieved trajectory accurately than state-of-the-art allowed proposed increase quality building compared approach by 10 percentage points. Our model had less value spatial coordinates Euler angles 25–35% 30–44%, respectively.

Language: Английский

---

Wildfire and power grid nexus in a changing climate

Nature Reviews Electrical Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: March 24, 2025

Language: Английский

---

Shift of emphasis toward intelligent equipment maintenance in port operations: A critical review of emerging trends and challenges

Ocean & Coastal Management, Journal Year: 2024, Volume and Issue: 259, P. 107408 - 107408

Published: Nov. 4, 2024

Language: Английский

---

A study on the detection of conductor quantity in cable cores based on YOLO-cable

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Dec. 28, 2024

The quantity of cable conductors is a crucial parameter in manufacturing, and accurately detecting the number can effectively promote digital transformation manufacturing industry. Challenges such as high density, adhesion, knife mark interference conductor images make intelligent detection particularly difficult. To address these challenges, this study proposes YOLO-cable model, which an improvement made upon YOLOv10 model. Specifically, Focal loss function introduced, C2F structure backbone optimized, NeXt module added, multi-scale feature (MSF) incorporated Neck section. Comparative experiments with various YOLO series models demonstrate that model significantly outperformed baseline YOLOv10s it achieves recall, mAP0.5, mAP scores 0.982, 0.994, 0.952, respectively. Further visualization analysis shows overlap boxes manually labeled samples reaches 90.9% length 95.7% height, indicating data consistency. IOU threshold adopted by enables to filter out false detection, thus ensuring accuracy. In short, proposed excels conductors, enhancing quality control production. This provides new insights technical support for application deep learning industrial inspections.

Language: Английский

---

SimMolCC: A Similarity of Automatically Detected Bio-Molecule Clusters between Fluorescent Cells

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(17), P. 7958 - 7958

Published: Sept. 6, 2024

In the field of studies on “Neural Synapses” in nervous system, its experts manually (or pseudo-automatically) detect bio-molecule clusters (e.g., proteins) many TIRF (Total Internal Reflection Fluorescence) images a fluorescent cell and analyze their static/dynamic behaviors. This paper proposes novel method for automatic detection image conducts several experiments performance, e.g., mAP @ IoU (mean Average Precision Intersection over Union) F1-score IoU, as an objective/quantitative means evaluation. As result, best proposed methods achieved 0.695 = 0.5 0.250 would have to be improved, especially with respect recall IoU. But, could automatically that are not only circular always uniform size, it can output various histograms heatmaps deeper analyses detected clusters, while particles by Mosaic Particle Tracker 2D/3D, which is one most conventional experts, size. addition, this defines validates similarity between cells, i.e., SimMolCC, also shows some examples SimMolCC-based applications.

Language: Английский

---

An Improved YOLOv8-Based Foreign Detection Algorithm for Transmission Lines

Sensors, Journal Year: 2024, Volume and Issue: 24(19), P. 6468 - 6468

Published: Oct. 7, 2024

This research aims to overcome three major challenges in foreign object detection on power transmission lines: data scarcity, background noise, and high computational costs. In the improved YOLOv8 algorithm, newly introduced lightweight GSCDown (Ghost Shuffle Channel Downsampling) module effectively captures subtle image features by combining 1 × convolution GSConv technology, thereby enhancing accuracy. CSPBlock (Cross-Stage Partial Block) fusion enhances model's accuracy stability strengthening feature expression spatial perception while maintaining algorithm's nature mitigating issue of vanishing gradients, making it suitable for efficient complex line environments. Additionally, PAM (pooling attention mechanism) distinguishes between target without adding extra parameters, even presence noise. Furthermore, AIGC (AI-generated content) technology is leveraged produce high-quality images training augmentation, lossless distillation ensures higher reduces false positives. conclusion, architecture parameter count 18% improving [email protected] metric a margin 5.5 points when compared YOLOv8n. Compared state-of-the-art real-time frameworks, our demonstrates significant advantages both model size.

Language: Английский

---