基于时空遥感特征耦合的农作物类型识别  

Research on crop type recognition based on spatio-temporal remote sensing feature coupling

作　　者：路颖 刘玉锋[1] 宋婷 汤鹤 LU Ying;LIU Yufeng;SONG Ting;TANG He(College of Geographic Information and Tourism,Chuzhou University,Chuzhou 239000,China;College of Economics and Management,Northeast Agriculfural University,Harbin 150030,China)

机构地区：[1]滁州学院地理信息与旅游学院,安徽滁州239000 [2]东北农业大学经济管理学院,哈尔滨150030

出　　处：《黑龙江工程学院学报》2021年第4期6-13,19,共9页Journal of Heilongjiang Institute of Technology

基　　金：高分专项省(自治区)域产业化应用项目(76-Y40G05-9001-15/18);国家级大学生创新创业训练计划项目(201910377057);滁州学院大学生创新创业训练计划资助项目(2019CXXL045)。

摘　　要：遥感技术凭借其能够精准地识别作物的分布情况以及动态监测作物的长势,已经被广泛应用到农业生产中。然而,传统的作物分类方法是基于单一中低分辨率遥感影像的像元进行分类,会存在地物混合、分类精度较低等问题。选择滁州市全椒县作为实验区,对GF1-PMS(2 m)与2018年10月至2019年9月内的11个时相的GF1-WFV(16 m)遥感影像进行预处理。运用时序特征分析、可分性比较分析、多尺度分割、面向对象等方法,提取作物分布信息并对分类后的结果进行精度评价。研究结果表明:采用中高分辨率卫星遥感影像相结合的方式,利用两者在农作物识别方面的优势进行农作物种植结构提取,作物识别的总体精度达到88.38%,效果显著。Remote sensing technology has been widely used in agricultural production by virtue of its ability to accurately identify the distribution of crops and dynamically monitor the growth of crops.However,the traditional crop classification method is based on the pixels of a single medium and low resolution remote sensing image,which has problems such as mixed ground objects and low classification accuracy.In this paper,Quanjiao County,Chuzhou City is selected as the experimental area to preprocess GF1-PMS(2m)and GF1-WVF(16m)remote sensing images in 11 time phases from October 2018 to September 2019.The time series feature analysis,separability comparative analysis,multi-scale segmentation,and object-oriented are used to extract crop distribution information and to evaluate the accuracy of the classified results.The research result shows that:using the combination of medium and high resolution satellite remote sensing images,and using the advantages of the two in crop identification to extract crop planting structure,the overall accuracy of crop identification reaches 88.38%,of which the effect is significant.

关 键 词：遥感 农作物类型识别 面向对象分类 NDVI 全椒县 

分 类 号：S127[农业科学—农业基础科学]