摘要
为对甜菜叶片氮含量进行快速估测,利用高光谱成像仪获取甜菜冠层叶片高光谱图像数据,通过凯氏定氮法测定叶片氮含量。基于精细采样法在全波段范围内构建归一化光谱指数(Normalized difference spectral index,NDSI)和土壤调节光谱指数(Soil-adjusted spectral index,SASI),并提出了基于粒子群算法的植被冠层调节参数L优化方法,探寻任意波段组合下SASI的最佳L值及其变化规律。在筛选出特征光谱指数基础上,开展甜菜叶片氮含量的定量估测和可视化研究。结果表明,各生育期SASI对甜菜冠层叶片氮含量(Canopy leaf nitrogen content,CLNC)的敏感度高于NDSI,尤其在NDSI易发生饱和现象的近红外区域。相比常规光谱指数,叶丛快速生长期基于SASI1(R430. 20,R896. 76)和SASI2(R433. 03,R896. 01)建立的CLNC估测模型预测效果最优,2015年验证集R^2为0. 78,RMSE为2. 48 g/kg,RE为4. 18%;糖分增长期以SASI3(R952. 09,R946. 11)和SASI4(R760. 37,R803. 48)的建模效果最佳,2015年验证集R^2为0. 67,RMSE为2. 71 g/kg,RE为4. 72%;糖分积累期的最优建模参数为SASI5(R883. 30,R887. 79),2015年模型R^2为0. 72,RMSE为2. 54 g/kg,RE为4. 49%。为直观显示甜菜CLNC在时间和空间尺度上的变化规律,基于上述估测模型计算并生成甜菜CLNC的预测分布图,实现了甜菜CLNC的可视化。研究结果表明,提出的甜菜CLNC估测方法具有可行性,可为及时了解作物长势及营养估测提供技术支持。
In order to estimate the nitrogen content of sugar beet leaves quickly,sugar beet was taken as the research object.The hyperspectral image data of canopy leaves was obtained by hyperspectral imaging spectrometer.At the same time,the nitrogen content of leaves was determined by Kjeldahl method.Based on the meticulous sampling method,the normalized spectral parameter(NDSI)and the soiladjusted vegetation index(SASI)were constructed in the whole-wavelength range.Moreover,in order to search for the optimal value of L in SASI under arbitrary band combination,the particle swarm optimization algorithm was proposed to optimize the L.On the basis of the previous work mentioned above,the sensitive spectral parameters were selected to achieve the optimization,and the estimation model was constructed to carry out the quantitative diagnosis and visualization research of the nitrogen content in the sugar beet leaves.The results indicated that the sensitivity of SASI to the canopy leaf nitrogen content(CLNC)of sugar beet was higher than that of NDSI for each different growth period.Especially in the near-infrared region where saturation easilyoccurred,the correlation was significantly improved.Compared with the conventional spectral parameters,based on SASI1(R430.20,R896.76)and SASI2(R433.03,R896.01),an optimal CLNC estimation model of BP net for the rapid growth period of the beet leaves was able to be established.The determination coefficient(R^2)of validation set was 0.78,the root mean square error(RMSE)was 2.48 g/kg and the relative error(RE)was 4.18%(in the year of2015).The model established based on SASI3(R952.09,R946.11)and SASI4(R760.37,R803.48)for the sugar growth period had the best performance.The R^2 of verification set was 0.67,the RMSE was 2.71 g/kg,and RE was 4.72%(in the year of 2015).The optimal modeling parameters for the sugar accumulation period were SASI5(R883.30,R887.79),and the R^2 of the model was 0.72,the RMSE was 2.54 g/kg,and the RE was4.49%(in the year of 2015).Based on the above model,combined with the spectral information of each band under every pixel of hyperspectral image,the CLNC was calculated,and the CLNC concentration graphs of sugar beet were plotted,which directly and visually presented the distribution of nitrogen content in the sugar beet leaves at different time scales and different leaf positions.The research results introduced that the proposed estimation method of CLNC in sugar beet was feasible,which also provided technical support for timely observation of crop growth and nutritional diagnosis.
作者
田海清
张晶
张珏
吴利斌
王迪
李斐
TIAN Haiqing;ZHANG Jing;ZHANG Jue;WU Libin;WANG Di;LI Fei(College of Mechanical and Electrical Engineering,Inner Mongolia Agricultural University,Huhhot 010018,China;College of Physics and Electronic Information,Inner Mongolia Normal University,Huhhot 010022,China;College of Grassland,Resources and Environment Science,Inner Mongolia Agricultural University,Huhhot 010019,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2019年第3期168-178,共11页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(41261084)
内蒙古自然科学基金项目(2016MS0346)
关键词
甜菜
叶片氮含量
高光谱图像
粒子群
光谱指数
sugar beet
leaf nitrogen content
hyperspectral image
particle swarm
spectral parameters
