大豆叶面积指数的高光谱估算模型研究  

Models for Estimating Soybean Leaf Area Index Using Hyperspectral Data

作　　者：黄春燕[1] 刘胜利[2] 王登伟[1] 战勇[2] 张恒斌[2] 袁杰[1] 马勤建[1] 陈燕[1] 赵鹏举[1] 

机构地区：[1]石河子大学新疆兵团绿洲生态农业重点实验室、石河子农学院,新疆石河子832003 [2]新疆农垦科学院作物所,新疆石河子832000

出　　处：《大豆科学》2008年第2期228-232,共5页Soybean Science

基　　金：新疆兵团“十一五”科技攻关项目资助(2006GG07)

摘　　要：通过测试大豆4个生育阶段350～2500nm波段的冠层高光谱数据,用近红外波段760nm～850nm及红光波段650nm～670nm的2个范围内的波段反射率,组成了高光谱比值植被指数(RVI)和800nm和670nm2个波段反射率组成修改型二次土壤调节植被指数(MSAVI2);基于RVI和MSAVI2植被指数,建立了大豆叶面积指数(LAI)的6种单变量线性与非线性函数模型,经检验均达到1%极显著水平。其中,以RVI所构建LAI的幂函数、MSAVI2所构建LAI的指数函数、对数函数估测模型的相关系数相对较高;用MSAVI2所构建的LAI精度较高的对数函数模型反演大豆叶面积指数,实测LAI与估测LAI呈极显著线性相关(R=0.9098**,n=46),模型方程的估算精度达84.9%,实测值与估算值的RMSE=0.2420,平均相对误差为0.1510。表明采用高光谱植被指数,能够实时、无损、动态、定量提取大豆叶面积指数,为设计理想的大豆群体和大豆遥感估产提供了科学的依据。Leaf area index（LAI） is an important parameter as the indicator of optimal diagnosis for crop growing status. Research shows that there are high correlation between hyperspectral data and LAI. So hyperspectral remote sensing can be used in monitoring growth status of soybean. In this paper, hyperspectral reflectance（350 to 2 500 nm） data was obtained in four soybean key growth stages, Ratio vegetation index（RVI） was computed using average reflectance of near infrared bands of 760 - 850 nm and red region bands of 650 -670 nm; Modified second soil-adjusted vegetation index（MSAVI2） was composed of reflectance of near infrared band of 800 nm and 670 nm. Based on RVI and MSAVI,six single variables of linear and nonlinear function models against LAI were established. All models reached 0.01 significance level, whilst ,power function fitting of RVI,exponential function fitting and logarithm function fitting of MSAV12 had comparatively higher accuracy for estimating soybean LAI; then the soybean canopy LAI was estimated according to the highest correlation coefficient of accurate logarithm model function between MSAVI2 and measured LAI,it showed that the correlation between measured LAI and estimated LAI was significant（R = 0. 9098 ^＊＊ , n =46 ）. The regression function accuracy was 84.9%, the RMSE was 0. 2420, average relative error was 0.1510. it is real-time, nondestructive and quantitative for adopting vegetation indices RVI, MSAVI2 to obtain soybean LAI,it can offer an evidence to design an optimum soybean canopy and estimate soybean yield by using hyperspectral remote sensing.

关 键 词：大豆 高光谱 叶面积指数 植被指数 估算模型 

分 类 号：S565.1[农业科学—作物学]