基于YOLOv8的航拍车辆检测技术研究

Research on Aerial Vehicle Detection Technology Based on YOLOv8

摘要: 随着经济和科技的发展，智慧交通逐渐被人们关注。作为智慧交通的关键技术，航拍车辆检测技术也逐渐走入人们的视野。针对目前航拍车辆目标较小，道路环境复杂造成的漏检率高的情况，本文提出了一种基于YOLOv8的航拍车辆检测方法。在YOLOv8的骨干网络中增加注意力机制模块，以增加网络间特征的联系，同时将头部网络的部分C2f层修改为C3层。经实验表明，改进后的网络在UCAS-AOD数据集上的准确率为93.4%，达到了不错的效果。

Abstract: With the development of economy and science and technology, intelligent transportation is gradu-ally being paid attention to. As the key technology of intelligent transportation, aerial vehicle detec-tion technology has also gradually come into people’s view. Aiming at the current situation that aerial vehicle targets are small and the road environment is complex resulting in a high rate of missed detection, this paper proposes an aerial vehicle detection method based on YOLOv8. The at-tention mechanism module is added to the backbone network of YOLOv8 to increase the linkage of features between networks, and part of the C2f layer of the head network is modified to C3 layer. It is experimentally shown that the improved network achieves a good accuracy of 93.4% on the UCAS-AOD dataset.