首页 | 本学科首页   官方微博 | 高级检索  
     检索      

深度学习遥感影像油罐检测算法精度对比分析
引用本文:李宸尧,郭海涛,马东洋,余东行,黄辰虎.深度学习遥感影像油罐检测算法精度对比分析[J].海洋测绘,2020,40(2):52-56.
作者姓名:李宸尧  郭海涛  马东洋  余东行  黄辰虎
作者单位:信息工程大学地理空间信息学院,河南郑州450001;61618部队,北京100094;海军研究院,天津300061
基金项目:国家自然科学基金(41671410)
摘    要:油罐目标的检测对于海洋战场环境保障具有重要的意义和作用。选择当前较为经典的几种深度学习目标检测算法,包括FRCNN、RFCN、SSD、YOLOv3、RetinaNet,利用已有的公开数据,对各算法进行油罐检测的精度进行深入对比分析和实验验证。实验结果表明,上述方法中鲁棒性和平均精度最好的是RFCN和RetinaNet;影像中目标的尺寸是影响各算法精度的重要因素。最后对基于深度学习的遥感影像油罐目标检测算法提出了改进的建议。相关研究对于利用深度学习算法完成油罐目标的实际检测应用具有重要的指导意义和参考价值。

关 键 词:遥感影像  目标检测  油罐检测  小目标检测  深度学习

Precision Analysis of Oil Tank Detection Algorithms Based on Deep Learning in Remote Sensing Image
LI Chenyao,GUO Haitao,MA Dongyang,YU Donghang,HUANG Chenhu.Precision Analysis of Oil Tank Detection Algorithms Based on Deep Learning in Remote Sensing Image[J].Hydrographic Surveying and Charting,2020,40(2):52-56.
Authors:LI Chenyao  GUO Haitao  MA Dongyang  YU Donghang  HUANG Chenhu
Institution:Institute of Geospatial Information,Information Engineering University,Zhengzhou 450001 ,China;61618 Troops,Beijing 100094 ,China; Naval Research Academy,Tianjin 300061 ,China
Abstract:The detection of oil tank is of great significance to the marine battlefield environment security. Several classical algorithms are selected for comparing,including FRCNN,RFCN,SSD,YOLOv3 and RetinaNet. Experiments and comparative analysis are completed on these algorithms by using public datasets. The results show that RFCN and RetinaNet have the best performance in average precision and robustness in condition of this experiment,and the main factor that affects the precision is the object size in the image. Finally,some suggestions for oil tank detection algorithm based on deep learning in remote sensing image are proposed. This research has great significance and value for application of oil tank detection based on deep learning.
Keywords:
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《海洋测绘》浏览原始摘要信息
点击此处可从《海洋测绘》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号