Image Detection Network Based on Enhanced Small Target Recognition Details and Its Application in Fine Granularity DOI Creative Commons
Qiang Fu, Xiaoping Tao,

Weijie Deng

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(11), P. 4857 - 4857

Published: June 4, 2024

Image detection technology is of paramount importance across various fields. This significance not only seen in general images with everyday scenes but also holds substantial research value the field remote sensing. Remote sensing involve capturing from aircraft or satellites. These typically feature diverse scenes, large image formats, and varying imaging heights, thus leading to numerous small-sized targets captured images. Accurately identifying these small targets, which may occupy a few pixels, challenging active area. Current methods mainly fall into two categories: enhancing target features by improving resolution increasing number bolster training datasets. However, approaches often fail address core distinguishing original images, resulting suboptimal performance fine-grained classification tasks. To this situation, we propose new network structure DDU (Downsample Difference Upsample), based on differential changing Neck layer deep learning networks enhance recognition further richness effectively solving problem low accuracy object recognition. At same time, order take account effect other sizes image, attention mechanism called PNOC (protecting channels) proposed, integrates universal without losing channels, thereby And experimental verification was conducted PASCAL-VOC dataset. it applied testing MAR20 dataset found that better than classic algorithms. because proposed framework belongs one-stage method, has good engineering applicability scalability, universality scientific applications are good. Through comparative experiments, our algorithm improved mAP 0.7% compared YOLOv8 algorithm.

Language: Английский

Visual fire detection using deep learning: A survey DOI
Guangtao Cheng, Xue Chen, Chenyi Wang

et al.

Neurocomputing, Journal Year: 2024, Volume and Issue: 596, P. 127975 - 127975

Published: June 1, 2024

Language: Английский

Citations

12
Improved YOLOv7 algorithm for flame detection in complex urban environments DOI
Qinghui Zhou,

Wuchao Zheng

Engineering Research Express, Journal Year: 2025, Volume and Issue: 7(1), P. 015283 - 015283

Published: March 17, 2025

Abstract To address the problems of flame detection, such as difficulties in detecting flames and poor performance complex urban environments, an improved YOLOv7-based detection algorithm for scenarios is proposed. The proposed increases multi-scale feature fusion introduces a 160 × scale, which improves capability small target flames. Additionally, 3 convolutions backbone extraction module YOLOv7 are replaced with deformable (Deformable Convolution Networks v2, DCNv2), better accommodate varying input map shapes enhance network’s learning ability scenarios. Furthermore, Convolutional Block Attention Module (CBAM) embedded to strengthen response relevant features, further improving algorithm’s dynamic environments. K-means++ used re-cluster anchor boxes, enhancing predict sizes locations. modified achieves mean Average Precision ([email protected]) 97.1%, improvement 4.9 percentage points. Experimental results demonstrate that significantly enhances

Language: Английский

Citations

0
SSOD-MViT: A novel model for recognizing alfalfa seed pod maturity based on semi-supervised learning DOI
Fuyang Tian, Y. Zhang, Shakeel Ahmed Soomro

et al.

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 236, P. 110439 - 110439

Published: April 23, 2025

Language: Английский

Citations

0
RT-DETR-Smoke: A Real-Time Transformer for Forest Smoke Detection DOI Creative Commons
Zhong Wang,

Lanfang Lei,

Tong Li

et al.

Fire, Journal Year: 2025, Volume and Issue: 8(5), P. 170 - 170

Published: April 27, 2025

Smoke detection is crucial for early fire prevention and the protection of lives property. Unlike generic object detection, smoke faces unique challenges due to smoke’s semitransparent, fluid nature, which often leads false positives in complex backgrounds missed detections—particularly around edges small targets. Moreover, high computational overhead further restricts real-world deployment. To tackle these issues, we propose RT-DETR-Smoke, a specialized real-time transformer-based smoke-detection framework. First, designed high-efficiency hybrid encoder that combines convolutional Transformer features, thus reducing cost while preserving details. We then incorporated an uncertainty-minimization strategy dynamically select most confident queries, improving accuracy challenging scenarios. Next, alleviate common issue blurred or incomplete boundaries, introduced coordinate attention mechanism, enhances spatial-feature fusion refines smoke-edge localization. Finally, WShapeIoU loss function accelerate model convergence boost precision bounding-box regression multiscale targets under diverse environmental conditions. As evaluated on our custom dataset, RT-DETR-Smoke achieves remarkable 87.75% [email protected] processes images at 445.50 FPS, significantly outperforming existing methods both speed. These results underscore potential practical deployment fire-warning smoke-monitoring systems.

Language: Английский

Citations

0
Intelligent Firefighting Technology for Drone Swarms with Multi-Sensor Integrated Path Planning: YOLOv8 Algorithm-Driven Fire Source Identification and Precision Deployment Strategy DOI Creative Commons

Bingxin Yu,

Shengze Yu,

Yuandi Zhao

et al.

Drones, Journal Year: 2025, Volume and Issue: 9(5), P. 348 - 348

Published: May 3, 2025

This study aims to improve the accuracy of fire source detection, efficiency path planning, and precision firefighting operations in drone swarms during emergencies. It proposes an intelligent technology for based on multi-sensor integrated planning. The integrates You Only Look Once version 8 (YOLOv8) algorithm its optimization strategies enhance real-time detection capabilities. Additionally, this employs data fusion swarm cooperative path-planning techniques optimize deployment materials flight paths, thereby improving precision. First, a deformable convolution module is introduced into backbone network YOLOv8 enable flexibly adjust receptive field when processing targets, enhancing accuracy. Second, attention mechanism incorporated neck portion YOLOv8, which focuses feature regions, significantly reducing interference from background noise further recognition complex environments. Finally, new High Intersection over Union (HIoU) loss function proposed address challenge computing localization classification targets. dynamically adjusts weight various components training, achieving more precise classification. In terms visual sensors, infrared LiDAR sensors adopts Information Acquisition Optimizer (IAO) Catch Fish Optimization Algorithm (CFOA) plan paths coordinated swarms. By adjusting planning locations, can reach sources shortest possible time carry out operations. Experimental results demonstrate that improves by optimizing algorithm, algorithms, strategies. optimized achieved 94.6% small fires, with false rate reduced 5.4%. wind speed compensation strategy effectively mitigated impact material deployment. not only enhances but also enables rapid response scenarios, offering broad application prospects, particularly urban forest disaster rescue.

Language: Английский

Citations

0
Low Complexity Forest Fire Detection Based on Improved YOLOv8 Network DOI Open Access
Lei Lin, Ruifeng Duan, Feng Yang

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(9), P. 1652 - 1652

Published: Sept. 19, 2024

Forest fires pose a significant threat to ecosystems and communities. This study introduces innovative enhancements the YOLOv8n object detection algorithm, significantly improving its efficiency accuracy for real-time forest fire monitoring. By employing Depthwise Separable Convolution Ghost Convolution, model’s computational complexity is reduced, making it suitable deployment on resource-constrained edge devices. Additionally, Dynamic UpSampling Coordinate Attention mechanisms enhance ability capture multi-scale features focus relevant regions, small-scale fires. The Distance-Intersection over Union loss function further optimizes training process, leading more accurate bounding box predictions. Experimental results comprehensive dataset demonstrate that our proposed model achieves 41% reduction in parameters 54% GFLOPs, while maintaining high mean Average Precision (mAP) of 99.0% at an Intersection (IoU) threshold 0.5. offers promising solution monitoring, enabling timely of, response to, wildfires.

Language: Английский

Citations

2
Image Detection Network Based on Enhanced Small Target Recognition Details and Its Application in Fine Granularity DOI Creative Commons
Qiang Fu, Xiaoping Tao,

Weijie Deng

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(11), P. 4857 - 4857

Published: June 4, 2024

Image detection technology is of paramount importance across various fields. This significance not only seen in general images with everyday scenes but also holds substantial research value the field remote sensing. Remote sensing involve capturing from aircraft or satellites. These typically feature diverse scenes, large image formats, and varying imaging heights, thus leading to numerous small-sized targets captured images. Accurately identifying these small targets, which may occupy a few pixels, challenging active area. Current methods mainly fall into two categories: enhancing target features by improving resolution increasing number bolster training datasets. However, approaches often fail address core distinguishing original images, resulting suboptimal performance fine-grained classification tasks. To this situation, we propose new network structure DDU (Downsample Difference Upsample), based on differential changing Neck layer deep learning networks enhance recognition further richness effectively solving problem low accuracy object recognition. At same time, order take account effect other sizes image, attention mechanism called PNOC (protecting channels) proposed, integrates universal without losing channels, thereby And experimental verification was conducted PASCAL-VOC dataset. it applied testing MAR20 dataset found that better than classic algorithms. because proposed framework belongs one-stage method, has good engineering applicability scalability, universality scientific applications are good. Through comparative experiments, our algorithm improved mAP 0.7% compared YOLOv8 algorithm.

Language: Английский

Citations

0