银杏落叶期叶片NDVI值2种拟合方法比较

Comparison of two fitting methods for NDVI of the deciduous period of Ginkgo leaves

【目的】对银杏Ginkgo biloba L.落叶期(9—12月)的叶片进行连续定期定点的高光谱测量,计算出能代表其生长状况和营养信息的高光谱参量NDVI值,通过选用两种不同的曲线拟合方法对其NDVI值进行宏观上的以时间为自变量的曲线拟合,选出最优拟合方法,更好地了解银杏落叶期叶片光谱特征参量NDVI值的变化趋势,从而更有效地对其进行决策和控制,为植被的大尺度遥感动态监测提供方法参考。【方法】利用SVC HR-1024I全波段地物光谱仪,选取三株健康、生长环境相同、长势相近的中龄银杏为叶片采集对象,对其落叶期冠层叶片进行定期定点定方位的高光谱观测。对获取的高光谱原始数据进行数据筛选与预处理后,通过计算得出叶片的NDVI值,分别采用二次函数拟合法和ARIMA时间序列拟合法对落叶期叶片的NDVI值进行曲线拟合,并对两种拟合方法的拟合结果进行比较,选出最适合银杏叶片落叶期NDVI值的拟合方法。【结果】二次函数拟合结果为NDVI=-0.0221T2+0.0547T+0.711,决定系数R2为0.926,但因拟合结果t值不显著,样本结果随机性大,不具广泛性;ARIMA时间序列拟合中ARIMA(2,1,2)模型估测结果与实际情况接近,R2为0.811,拟合效果较好。【结论】ARIMA时间序列拟合方法比二次行数拟合方法更适用于对银杏落叶期叶片的NDVI值进行拟合。

【Objective】The NDVI parameters of the leaves of Ginkgo biloba leaves were fitted with a curve with macroscopic time as the independent variable,it is possible to understand the changing trend of the spectral characteristics of leaves in the leaves of Ginkgo biloba,so as to make decision and control,and provide reference for the large-scale remote sensing dynamic monitoring of vegetation.【Method】Use the SVC HR-1024I Full-Band ASD Spectrometer to observe the hyperspectral data of Ginkgo’s leaves during it’s deciduous period(Sep-Dec)at a fixed time and location,then process the hyperspectral data and calculate NDVI value.Separately carry out the least square method of quadratic function fitting and ARIMA time series fitting method to find out the best fitting method by comparing.【Result】The quadratic function fitting result is NDVI=-0.0221T2+0.0547T+0.711,The decision coefficient R2 is 0.926,but the t value of the fitting result is not significant,this shows that the fitting result is random and nonrepresentative;ARIMA time sequence fitting results are ARIMA(2,1,2)model,R2=0.811,and the prediction trend is roughly the same as the actual trend.【Conclusion】The results show that the ARIMA time series is more suitable for the fitting of the NDVI value of leaf blade.