基于改进YOLOv8的无人机航拍图像目标检测算法
Object Detection Algorithm for UAV Aerial Image Based on Improved YOLOv8
程换新 1乔庆元 1骆晓玲 2于沙家1
作者信息
- 1. 青岛科技大学 自动化与电子工程学院,山东青岛 266061
- 2. 青岛科技大学机电工程学院,山东青岛 266061
- 折叠
摘要
针对现存无人机航拍图像目标检测算法检测精度较低、模型较为复杂的问题,提出一种改进YOLOv8的目标检测算法.在骨干网络引入多尺度注意力EMA,捕捉细节信息,以提高模型的特征提取能力;改进C2f模块,减小模型的计算量.提出了轻量级的Bi-YOLOv8特征金字塔网络结构改进YOLOv8的颈部,增强了模型多尺度特征融合能力,改善网络对小目标的检测精度.使用WIoU Loss优化原网络损失函数,引入一种动态非单调聚焦机制,提高模型的泛化能力.在无人机航拍数据集VisDrone2019上的实验表明,提出算法的mAP50为40.7％,较YOLOv8s提升了 1.5％,参数量降低了42％,同时相比于其他先进的目标检测算法在精度和速度上均有提升,证明了改进算法的有效性和先进性.
Abstract
To solve the problem that the existing UAV aerial image target detection algorithm has low detection accuracy and complex model,an improved YOLOv8 target detection algorithm is proposed.Multi-scale attention EMA is introduced into the backbone network to capture detailed information to improve the feature extraction ability and C2f module is improved to reduce the calculation amount of the model.The lightweight Bi-YOLOv8 feature pyramid network structure is proposed to improve the neck of YOLOv8,the multi-scale feature fusion ability of the model is enhanced,and the detection accuracy of the network for small targets is improved.WIoU Loss is used to optimize the original network loss function,and a dynamic non-monotonic focusing mechanism is introduced to improve the generalization ability of the model.Experiments on UAV aerial image data set VisDrone2019 show that the mAP50 of the proposed algorithm is 40.7％,which is 1.5％higher than YOLOv8s,and the number of parameters is reduced by 42％.The accuracy and speed are improved compared with other advanced target detection algorithms,which proves the effectiveness and advanced nature of the proposed algorithm.
关键词
航拍图像/小目标检测/YOLOv8/Bi-YOLOv8/轻量化Key words
aerial images/small object detection/YOLOv8/Bi-YOLOv8/lightweight引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据