基于GF-1与Landsat-8多光谱遥感影像的玉米LAI反演比较  

Comparison between GF-1 images and Landsat-8 images in monitoring maize LAI

作　　者：贾玉秋 李冰 程永政[1] 刘婷[1] 郭燕[1] 武喜红[1] 王来刚[1] 

机构地区：[1]河南省农业科学院农业经济与信息研究所,郑州450002 [2]郑州澍青医学高等专科学校,郑州450064

出　　处：《农业工程学报》2015年第9期173-179,共7页Transactions of the Chinese Society of Agricultural Engineering

基　　金：高分辨率对地观测系统重大专项项目(09-Y30B03-9001-13/15);河南省科技成果转化项目(14220111033);河南省农业科学院农业科技创新项目(201315618)

摘　　要：近年来,中国遥感事业已取得长足进步,高分一号(GF-1)卫星首次实现了中国自主研发的高分辨率对地观测。为探讨国产GF-1卫星影像在农业遥感长势监测中的适应性,以许昌地区为研究对象,选取同期Landsat-8卫星影像,结合地面采样数据LAI,从传感器光谱响应特征、经验回归模型监测精度以及LAI空间一致性等3方面进行2类遥感数据的对比评价。结果表明,GF-1影像近红外、红、蓝波段光谱响应与Landsat-8有差异,与绿波段光谱响应非常吻合,各波段光谱反射率与Landsat-8影像同类光谱间均存在显著线性关系。通过各波段组合多种归一化植被指数,采用经验回归模型反演LAI发现,GF-1影像反演的最优模型为NDVI的指数模型,R2为0.848,Landsat-8影像反演的最优模型为蓝红组合的归一化植被指数(blue-red NDVI,BRNDVI)的指数模型,R2为0.687,2类影像反演LAI与地面实测值均呈现较为一致的线性关系。由许昌地区玉米LAI值空间分布可见,GF-1影像反演的玉米LAI值与Landsat-8影像反演值过渡趋势一致,在许昌西部种植结构复杂地区,GF-1影像以其空间分辨率优势更能凸显LAI分布差异。通过该文研究表明,GF-1卫星的高时间分辨率以及高空间分辨率特征能够代替传统中分辨率数据成为农业遥感长势监测中的重要数据源,该数据在农业遥感其他领域的应用是今后研究的重点。With China Remote Sensing career advancement, a large number of independent researches and development of satellite have launched. Among a new generation of high-resolution satellites, GF-1 stands out. It sets high spatial resolution, multi-spectral and high temporal resolution in a fusion technology with strategic significance. To explore Chinese GF-1 satellite images’ adaptability of agricultural growth monitoring, its images for the region of Xuchange China for maize growth were compared with the same period of Landsat-8 satellite images in three aspects of sensor spectral response characteristics, the accuracy of empirical regression model and LAI space consistency. There were a total of 24 sampling points for the study. First, graphs described the sample located pixels’ spectral reflectance of near-infrared band, red band, green band and blue band of the two types of sensors. It directly reflected the spectral reflectance differences between sensors in the same place, and differences between maize in different area. The reflectance of near-infrared and red band of Landsat-8 was higher compared with GF-1. The blue and green band’s reflectance of GF-1 was similar to that of Landsat-8. The linear correlation of two sensors’ reflectivity could be calculated at the same time. Second, four bands of two types of images were separately combined into seven kinds of normalized difference vegetation index to further eliminate the influence of atmospheric correction process. Like NDVI, the red band was replaced by blue or green or three visible bands’ combination of two by two or sum of them. Then, the empirical regression models were used to calculate the ability of inversing LAI among the vegetation index. Based on comparison ofR2 and RMSE among models, high fitting models were selected. The optimal model for Landsat-8 was based on BRNDVI, it was an index model. The best model for GF-1was based on NDVI, and model type was an index model. The reserved samples were used to test model’s fitting accuracy. The

关 键 词：遥感 作物 波长 玉米 GF-1 Landsat-8 叶面积指数 

分 类 号：S127[农业科学—农业基础科学]