摘要
【目的】研究棉花冠层光谱对不同叶面积指数(LAI)的响应,建立棉花LAI光谱反演模型。【方法】利用2003～2004年采集的棉花光谱与LAI的246组数据,分析LAI与冠层反射率光谱和反射率一阶微分光谱间的定量关系。【结果】当LAI大于2.5后不同LAI棉花群体光谱反射率在可见光波段趋于饱和;LAI与可见光波段和短波红外波段(水分吸收带除外)光谱反射率呈显著负相关,与近红外波段高光谱反射率呈显著正相关;LAI与棉花反射率一阶微分光谱主要在蓝边(523～531nm)、黄边(570～576nm)、红边(700～755nm)形成3个相关系数高台区,均达极显著水平,其中红边区的相关性最高。棉花红边位置固定,分别在718nm和723nm,且以723nm处对LAI更敏感。在反演棉花LAI的高光谱参数中VI(660、800)、VI(550、800)、VI(500、800)、VI(670、800)、Sdy(570～573nm)、SDr(714～755nm)、D723、Dr估算LAI相对误差低于30%,RSME小于0.6,其中VI(600、800)、VI(550、800)两个参数估算水平最高,相对误差分别为21.7%与21.0%,RMSE分别为0.416与0.419;利用SDr与SDr/SDb分别对LAI大于1.0与小于1.0的棉花群体反演,能显著提高LAI的估算水平。【结论】应用高光谱分析方法能够提取棉花冠层特征光谱信息,构建LAI高光谱反演参数,建立估算模型,并且利用包含不同光谱参数的分段模型可以进一步提高LAI反演精度。
[Objective] In order to establish the retrieval models based on spectrum in cotton, the spectrum response on the cotton canopy under the different leaf area index were researched. [Method] The relationship between LAI and reflectance and the first order different value was analyzed with the 246 couple datum of the spectrum and LAI collected in the 2003-2004 year. [Result] The results showed that the reflectance difference in the visional waveband with the cotton canopies of the different LAI tend to the zero after 2.5 of LAI; the correlation between LAI and the spectrum is significant negative in the visional waveband, and the correlation between LAI and the near infrared is significant positive; there are three high plaint regions, including blue region（523-531 nm）、 yellow region（570-576 nm）, red region（700-755 nm） respectively, resulted from the correlation between LAI and the first order different spectrum, and among them the red region is the strongest. The position of the red edge is fixation, located at 718 nm and 723 nm respectively, and importantly the reflectance at 723 nm is more sensitive to the leaf area index. Among hyperspectral parameters constructing the estimation model of LAI of cotton, the analyses shows that the relative error, including VI （660, 800）, VI （550, 800）, VI（500, 800）, VI（670, 800）, SDy（570-573 nm）, SDr（714-755 nm）, D723, Dr, is under 30% and RMSE is under 0.6, and VI （660, 800） and VI （550, 800） are better comparatively. The achievement is improved throughout establishing the subsection models, and SDr is used when the leaf index is under 1.0 and SDr/SDb is used when the leaf area index is above 1.0. [ Conclusion] In the research, the hyperspectral analyzing methods were used to attain the characteristic spectrum information in cotton. As a result, the retrieval parameters were constructed and the estimation models were established, the subsection models could further improve the retrieval precision of LAI with the different spectrum
出处
《中国农业科学》
CAS
CSCD
北大核心
2007年第1期63-69,共7页
Scientia Agricultura Sinica
基金
新疆兵团博士基金
国家"863"(20060110Z2053)
绿洲生态农业重点实验室开放课题资助项目
