摘要
针对小型低空飞行器在复杂环境中检测难的问题,提出了一种基于改进YOLOv8n模型的视觉检测方法。通过引入轻量级的AKConv卷积模块、SEAM通道空间混合域注意力模块和排斥损失函数,构建了YOLOv8-SE模型,并利用自制数据集进行训练与测试。AKConv模块通过动态调整卷积核的采样位置提升了特征提取的灵活性,SEAM模块增强了关键特征的捕捉能力,排斥损失函数则改善了遮挡环境下的检测精度。实验结果表明,相较于传统模型,YOLOv8-SE在多个评价指标上均表现优异。该研究为低空飞行器检测领域提供了一种高效、可靠的解决方案,为资源受限设备上的深度学习模型优化提供了新思路。
This study addresses the challenge of detecting small low-altitude aircraft in complex environments,and proposes a visual detection method based on an improved YOLOv8n model.The YOLOv8-SE model was constructed by integrating the lightweight AKConv convolution module,the SEAM channel-spatial mixed-domain attention module,and the repulsion loss function.A custom dataset was used for training and testing.The AKConv module enhances the flexibility of feature extraction by dynamically adjusting the sampling positions of convolution kernels,the SEAM module improves the ability to capture key features,and the repulsion loss function enhances detection accuracy in occluded environments.Experimental results demonstrate that YOLOv8-SE outperforms traditional models across multiple evaluation metrics.This research provides an efficient and reliable solution for small aerial targets detection and offers new insights into optimizing deep learning models for resource-constrained devices.
作者
杨光飞
张蕾蕾
Yang Guangfei;Zhang Leilei(Institute of Systems Engineering,Dalian University of Technology,Dalian 116024,China;Institute of Advanced Intelligence,Dalian University of Technology,Dalian 116024,China)
出处
《网络安全与数据治理》
2025年第12期39-47,共9页
CYBER SECURITY AND DATA GOVERNANCE
基金
国家自然科学基金(42071273)
中央高校基本科研业务费项目(DUT24YG147)。
