无人机影像反演玉米冠层LAI和叶绿素含量的参数确定

Determining the retrieving parameters of corn canopy LAI and chlorophyll content computed using UAV image

小型低空无人机(Unmanned Aerial Vehicle,UAV)机动灵活、操作简便,可以按需获取高空间分辨率影像,是育种玉米长势监测的一种重要技术手段。针对UAV影像反演玉米冠层叶面积指数(LAI,Leaf Area Index)和叶绿素含量的参数确定问题,该研究以DJI S1000+无人机为平台,搭载法国Parrot Sequoia相机,获取海南三亚市崖城玉米育种基地的多光谱影像。基于预处理后的UAV影像,采用重采样的方式获得不同分辨率下(0.1~1 m)的不同植被指数,所构建的植被指数包括归一化植被指数(Normalized Difference Vegetation Index,NDVI)、叶绿素指数(Grassland Chlorophyll Index,GCI)、比值植被指数(Ratio Vegetation Index,RVI)、归一化红边红指数(Normalized Difference rededge-red Index,NDIrer)、归一化红边绿指数(Normalized Difference rededge-green Index,NDIreg)和重归一化植被指数(Renormalized Difference Vegetation Index,RDVI),通过将不同分辨率下的不同植被指数与地面实测数据进行回归分析,以获得各分辨率下植被指数与冠层LAI和叶绿素含量的关系模型及其决定系数,以决定系数的大小为依据来确定玉米冠层LAI和叶绿素含量反演的最优空间分辨率和最优植被指数。通过试验发现,在分辨率为0.6 m时,NDVI与地面实测LAI之间的决定系数R^2为0.80,决定系数达到了最大,利用该分辨率下的NDVI反演得到的LAI验证精度R^2达到0.73;在分辨率为0.1 m时,NDIreg与地面实测叶绿素含量之间的决定系数R^2为0.70,决定系数达到最大,利用该分辨率下的NDIreg反演得到的叶绿素含量验证精度R^2达到了0.63。因此得出结论:1)植被指数的选择:①对于玉米冠层LAI的反演来说,不包含绿波段的植被指数的LAI反演精度较高,这说明绿波段对LAI的变化不敏感;②对于玉米冠层叶绿素含量反演来说,包含红边波段的植被指数的反演精度较高,因此影像的红边波段对叶绿素含量的变化非常敏感。2)UAV影像空间分辨率的选择:反演LAI的最优分辨率是0.6 m,此时NDVI与实测LAI的决定系数达到最大;反演冠层叶绿素含量的最优分辨率是0.1~0.3 m范围内,此时NDIreg与实测叶绿素含量的决定系数达到最大。该研究可为UAV反演玉米表型参数时的分辨率和植被指数选择提供参考。

The small low-altitude unmanned aerial vehicle(UAV)is flexible and easy to operate,which can be used to acquire high spatial resolution images with centimeter level.It is an important technical way for phenotyping the breeding corn.In order to determine the retrieving parameters of the corn canopy leaf area index(LAI)and chlorophyll content computed by UAV images,the DJI S1000+UAV platform with the French Parrot Sequoia camera was used to obtain multispectral images in Yacheng corn breeding base,Sanya City,Hainan Province in this study.Six different kinds of vegetation indices were used in computing the corn canopy LAI and chlorophyll content,each vegetation index was obtained from images with 10 spatial resolutions ranging from 0.1 to 1 m.The vegetation indices used in this study were the normalized difference vegetation index(NDVI),grassland chlorophyll index(GCI),ratio vegetation index(RVI),normalized difference rededge-red index(NDIrer),normalized difference rededge-green index(NDIreg)and renormalized difference vegetation index(RDVI).The correlation analysis between different vegetation indices from different resolutions images and in-situ measured LAI was done to select the optimal spatial resolution and optimal vegetation index for computing corn canopy LAI,and in similar for chlorophyll content.The study results revealed that the NDVI from image with 0.6 m spatial resolution was the optimal selection for LAI computing,where the correlation coefficient R^2 between NDVI and the in-situ measured LAI was 0.80 with a R^2 of 0.73 for verification.And the highest correlation coefficient R^2 between NDIreg from image with 0.1m spatial resolution and in-situ measured chlorophyll content is 0.70,with a R^2 of 0.63 for verification.The conclusions of this study were as followed:1)The selection of vegetation index:①For the corn canopy LAI computing,the vegetation indices without green band were higher than that with green band,which revealed that the green band was not sensitive to LAI;②For corn canopy chlorophyll content computing,the vegetation indices including the red edge band were higher than that without red edge band,which revealed that the red edge band was very sensitive to chlorophyll content.2)The spatial resolution selection for UAV image:the optimal resolution for LAI computing was 0.6 m when the correlation coefficient between NDVI and measured LAI reached the maximum;the optimal resolution for canopy chlorophyll content computing was 0.1-0.3 m when the correlation coefficient between NDIreg and in-situ measured chlorophyll content reached the maximum.This study can be used to give the reference for spatial resolution selection and vegetation index selection for corn canopy LAI and chlorophyll content computing by UAV image.