A Small-Object Detection Model Based on Improved YOLOv8s for UAV Image Scenarios
Remote Sensing · 05.07.2024 · 16 (13) · 2465
DOI: 10.3390/rs16132465 ↗Kurz gefasst
Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 05.07.2024. Verfasst von Jianjun Ni und 4 weiteren Autor:innen. Die Arbeit wurde bisher 86-mal zitiert (Crossref) und nennt 65 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Jiangsu Province Key R&D Program.
Abstract (Originalsprache)
Small object detection for unmanned aerial vehicle (UAV) image scenarios is a challenging task in the computer vision field. Some problems should be further studied, such as the dense small objects and background noise in high-altitude aerial photography images. To address these issues, an enhanced YOLOv8s-based model for detecting small objects is presented. The proposed model incorporates a parallel multi-scale feature extraction module (PMSE), which enhances the feature extraction capability for small objects by generating adaptive weights with different receptive fields through parallel dilated convolution and deformable convolution, and integrating the generated weight information into shallow feature maps. Then, a scale compensation feature pyramid network (SCFPN) is designed to integrate the spatial feature information derived from the shallow neural network layers with the semantic data extracted from the higher layers of the network, thereby enhancing the network’s capacity for representing features. Furthermore, the largest-object detection layer is removed from the original detection layers, and an ultra-small-object detection layer is applied, with the objective of improving the network’s detection performance for small objects. Finally, the WIOU loss function is employed to balance high- and low-quality samples in the dataset. The results of the experiments conducted on the two public datasets illustrate that the proposed model can enhance the object detection accuracy in UAV image scenarios.
Förderung
- National Natural Science Foundation of China
- Jiangsu Province Key R&D Program
Zitieren
APA
Ni, J., Zhu, S., Tang, G., Ke, C., & Wang, T. (2024). A Small-Object Detection Model Based on Improved YOLOv8s for UAV Image Scenarios. Remote Sensing, 16(13), 2465. https://doi.org/10.3390/rs16132465
Vancouver
Ni J, Zhu S, Tang G, Ke C, Wang T. A Small-Object Detection Model Based on Improved YOLOv8s for UAV Image Scenarios. Remote Sensing. 2024;16(13):2465. doi:10.3390/rs16132465
BibTeX
@article{Ni2024,
author = {Ni, Jianjun and Zhu, Shengjie and Tang, Guangyi and Ke, Chunyan and Wang, Tingting},
title = {A Small-Object Detection Model Based on Improved YOLOv8s for UAV Image Scenarios},
journal = {Remote Sensing},
year = {2024},
volume = {16},
number = {13},
pages = {2465},
doi = {10.3390/rs16132465}
}
EUVIMED