基于SIOU函数的改进YOLOv5遥感目标检测方法  

Improved YOLOv5 Remote Sensing Target Detection Method Based on SIOU Function

作　　者：高倩 潘杨[1] 朱磊[1] 杨金花 GAO Qian;PAN Yang;ZHU Lei;YANG Jinhua(Xi'an Polytechnic University,School of Electronics and Information,Xi'an 710048,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《长江信息通信》2022年第11期5-8,共4页Changjiang Information & Communications

摘　　要：为提升YOLOv5算法对遥感图像密集目标的检测精度并改善漏检问题,提出了一种改进的YOLOv5遥感目标检测算法。改进方法首先采用7*7卷积模块替换骨干网络中Focus模块以增大模型感受野;其次,在保证与原模块效果相同的情况下使用SPPF以提升检测速度;最后,引入SIOU损失函数,利用边界框回归之间的向量角度来重新定义损失函数,有效提高了检测的准确性。实验结果表明,针对公开的NWPU VHR-10遥感数据集,所提改进算法在保持与原算法相同检测速度的情况下,检测精度提高了3.5%。In order to improve the detection accuracy of YOLOv5 algorithm for dense targets in remote sensing images and improve the problem of missed detection, an improved YOLOv5 remote sensing target detection algorithm is proposed. The improved method firstly replaces the Focus module in the backbone network with a 7*7 convolution module to increase the receptive field of the model;secondly, SPPF is used to improve the detection speed while ensuring the same effect as the original module;finally, the SIOU loss function is introduced, using The vector angle between bounding box regressions is used to redefine the loss function, which effectively improves the detection accuracy. The experimental results show that for the public NWPU VHR-10 remote sensing data set, the proposed improved algorithm can improve the detection accuracy by 3.5% while maintaining the same detection speed as the original algorithm.

关 键 词：目标检测 遥感图像 SIOU YOLOv5 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]