| 本文已被：浏览 852次 下载 320次 |
码上扫一扫！ |
| 县域尺度上基于GF-1 PMS影像的冬小麦种植面积遥感监测 |
|
田海峰1,2, 周伯燕3, 陈燕芬1,2, 邬明权1, 牛铮1,2
|
|
|
| (1.中国科学院 遥感与数字地球研究所 遥感科学国家重点实验室, 北京 100101;2.中国科学院大学 环境与资源学院, 北京 100049;3.河南大学 环境与规划学院, 河南 开封 475004) |
|
| 摘要: |
| 为探究县域尺度上基于高分一号卫星（GF-1） PMS影像进行冬小麦遥感监测的可行性及精准性，以河南省滑县为研究区，遴选2015年2月上旬GF-1 PMS影像6景，对影像进行辐射定标、FLAASH大气校正、NNDiffuse融合、几何精校正、地图投影转换等预处理后，在外业调查和样本分析的基础上构建一种新的冬小麦决策树分类模型，模型第1层决策方案中NDVI>0.311的像元为冬小麦，得到冬小麦的粗分类结果；在此基础上进行第2层决策分类，以进一步提高冬小麦的分类精度，分类方案为第1波段地表反射率<0.146、第2波段地表反射率<0.148、第3波段地表反射率<0.135、第4波段地表反射率>0.250的像元为冬小麦。对分类结果进行形态学滤波处理，以消除或减少分类结果中孤立的像元。分别基于决策树分类模型与ENVI软件自带的IsoData非监督分类模型，对比分析GF-1 PMS影像和同时期Landsat-8 OLI影像在冬小麦面积提取上的精度。结果表明：基于新构建的决策树分类模型，2015年滑县冬小麦种植面积为115 715.81 hm2，混淆矩阵检验总体精度为99.62%，Kappa系数为0.99；PMS影像提取冬小麦的混淆矩阵总体精度比OLI影像高出9个百分点。说明县域尺度上基于单时相GF-1 PMS影像在冬小麦收获前提取冬小麦种植面积是可行的，提取精度较高。 |
| 关键词: 冬小麦 高分一号 县域尺度 决策树 PMS影像 Landsat |
| DOI：10.11841/j.issn.1007-4333.2017.10.17 |
| 投稿时间：2016-09-19 |
| 基金项目:国家自然科学基金项目（41301390，4137138）；国家"973"计划项目（2013CB733405）；国家"863"计划项目（2014AA06A511）；云南省科技计划（2010AD004）；高分辨率国家重大专项（20-Y30B17-9001-14/16） |
|
| Extraction of winter wheat acreage based on GF-1 PMS remote sensing image on county scale |
|
TIAN Haifeng1,2, ZHOU Boyan3, CHEN Yanfen1,2, WU Mingquan1, NIU Zheng1,2
|
| (1.The State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China;2.College of Resource and Environment, University of Chinese Academy of Sciences, Beijing 100049, China;3.College of Environment and Planning, Henan University, Kaifeng 475004, China) |
| Abstract: |
| The aim of this study was to understand the feasibility and accuracy of monitoring winter wheat acreage based on the GF-1 PMS image of county scale by taking Huaxian County of Henan Province as study area.Six scenes of GF-1 PMS image in early February 2015 were selected.The pretreatment work included radiometric calibration,FLAASH atmospheric correction,NNDiffuse Pan Sharpening,image geometry registration,map projection transformation,etc..Based on field survey data and sample data,a new decision tree model for winter wheat classification was constructed.The pixel was considered to winter wheat if its NDVI value is more than 0.311 in the level 1 decision scheme.However,this decision scheme was not very accurate for winter wheat classification.In order to improve the classification accuracy,the level 2 decision scheme was designed that the pixel was considered to winter wheat if its first band was less than 0.146,second band was less than 0.148,third band was less than 0.135,and fourth band was more than 0.250.The morphological filter was performed in the classification results in order to eliminate or reduce the isolated cells in classification results.The result showed that the winter wheat acreage based on our established decision tree model was 115 715.81 hm2 in 2015,overall accuracy was 99.62%,and Kappa coefficient was 0.99.The accuracy of extraction winter wheat based on PMS images was better than that based on Landsat-8 OLI images (more than 9%).This study proved that it was feasible and accurate to extract winter wheat acreage based on GF-1 PMS imagery on county scale.This study will provided technical support for monitoring winter wheat on the county scale based on high resolution remote sensing data. |
| Key words: winter wheat GF-1 county scale decision tree PMS imagery Landsat |
