电光与控制, 2023, 30 (6): 60, 网络出版: 2023-11-29  

基于改进YOLO模型的遥感小目标检测

Remote Sensing Small Target Detection Based on Improved YOLO Model
王凯 1,2王伟 1,2蒋志伟 1,2
作者单位
1 河北工业大学电子信息工程学院, 天津 300000
2 天津市电子材料与器件重点实验室, 天津 300000
摘要
针对遥感图像中小目标检测精度低以及漏检现象严重的问题, 提出一种基于YOLOv4改进的遥感小目标检测算法。该算法首先改进特征提取网络, 删除深层次特征层, 减少语义丢失现象;其次将轻量级注意力机制与RFB-S结构融合, 拓展感受野, 并加强网络对重要信息的关注程度, 从而提升检测精度;最后使用Focal Loss函数解决正负样本不均衡问题, 抑制背景目标, 进一步增强检测效果。在RSOD数据集上的实验结果表明, 改进后算法检测平均精度为96.5%, 召回率达到87.2%, 检测效果明显提升, 有效改善了小目标漏检现象, 对遥感图像小目标检测具有重要意义。
Abstract
A remote sensing small target detection algorithm based on the improved YOLOv4 is proposed to address the problems of low detection accuracy and serious missing detection of small targets in remote sensing images.Firstly, the feature extraction network is improved by removing the deep feature layer to reduce semantic loss.Secondly, the lightweight attention mechanism is fused with the RFB-S structure to expand the perceptual field and enhance attention to important information, thus improving the detection precision.Finally, the Focal Loss is used to avoid the imbalance between positive and negative samples and suppress the background targets to further enhance the detection effect.The experimental results on the RSOD dataset show that the improved algorithm has an average detection precision of 96.5% and a recall rate of 87.2%, which significantly improves the detection effect and effectively avoids the phenomenon of small target miss detection, and is of great significance to small target detection in remote sensing images.

王凯, 王伟, 蒋志伟. 基于改进YOLO模型的遥感小目标检测[J]. 电光与控制, 2023, 30(6): 60. WANG Kai, WANG Wei, JIANG Zhiwei. Remote Sensing Small Target Detection Based on Improved YOLO Model[J]. Electronics Optics & Control, 2023, 30(6): 60.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!