采用深度学习的遥感图像花生种植区域分类技术研究  

Research on Peanut Planting Area Classification Techonology Using Remote Sensing Image Based Deep Learning

作　　者：黄云 唐林波[1,2] 李震[1,2] 龙腾 Huang Yun;Tang Linbo;Li Zhen;Long Teng(Beijing Key Laboratory of Embedded Real-time Information Processing Technology,Beijing Institute of Technology,Beijing100081,China;School of Information and Electronics,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]北京理工大学嵌入式实时信息处理技术北京市重点实验室,北京100081 [2]北京理工大学信息与电子学院,北京100081

出　　处：《信号处理》2019年第4期617-622,共6页Journal of Signal Processing

基　　金：高等学校学科创新引智计划资助(B14010);“长江学者奖励计划”(T2012122)

摘　　要：近年来,深度学习在图像处理和数据分析等方面取得了巨大的进展。针对传统遥感估计农作物种植面积统计方法时效性差、依赖人工操作经验、耗费人力资源等问题,以Sentinel-2卫星遥感影像为数据基础,提出了一种基于深度学习的农作物种植区域分类方法。实验以从背景中提取出花生种植区域为目标,首先对Sentinel-2遥感影像数据进行预处理,然后用人工目视解译的方法标注遥感影像中种植花生的区域,将标注后的图像输入到图像分割网络中进行训练,最后将测试图像输入到训练好的分割网络,获得测试结果:检测准确率为89.20%,检测召回率为79.22%。In recent years,deep learning has made great progress in many fields,such as image processing and data analysis.The current fine extraction of crop areas relies mainly on computer-assisted manual visual interpretation.Since there are many disadvantages in crop planting area estimation using method of traditional remote sensing,which are poor timeliness,relying on manual operation experience and high human resource consumptions,a remote sensing crop planting area classification method based on deep learning was proposed using Sentinel-2 satellite remote sensing image.In order to obtain the peanut planting area,the Sentinel-2 remote sensing image data was firstly preprocessed,and then the peanut planting part was manually labeled to get the training and test data set.The training data set was then input into the image segmentation network for training.After training,the test data set was input into the network to obtain the test result:the detection accuracy rate was 89.20%,and the detection recall rate was 79.22%.

关 键 词：农作物 分类 深度学习 遥感 农业监测 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]