两种NDVI时间序列数据拟合方法比较

Comparison of two fitting methods of NDVI time series datasets

归一化植被指数(NDVI)时间序列数据拟合目的是降低时序数据的噪声水平,重建高质量的NDVI时序数据,有利于多种参数反演和信息提取。针对国际上普遍应用的两种NDVI时间序列数据拟合方法,即Savitzky-Golay滤波法和非对称性高斯函数拟合法,该文在介绍两种方法基本概念的基础上,利用直接比较法和间接比较法在中国对两种拟合方法进行了比较分析。结果表明,Savitzky-Golay滤波法和非对称性高斯函数拟合法的拟合效果总体上一致,但二者之间还是存在区域差异性,这种区域差异与两种方法的自身特点和中国区域自然条件紧密相关。不同数据拟合方法的比较研究可以弄清每种方法的优缺点和区域适宜性,有助于研究人员针对不同研究目的和研究区域选择适宜的NDVI数据拟合方法,减少遥感数据处理中的误差,提高研究精度。

The fitting of NDVI time series datasets is to minimize the effects of anomalous values caused by atmospheric haze and cloud contamination, and to reconstruct high-quality NDVI data for parameters inversion and information extraction. A comparative study on two well-known fitting approaches of NDVI time series data, namely, Savitzky-Golay filtering and asymmetric Gaussian function fitting, was described in China using both direct and indirect comparison methods. The results showed that these two methods generally presented a similar performance and a high agreement in data filtering in China, but regional differences existed between them. The main reason for their discrepancies was related to the methodological differences between these two methods, as well as the landscape heterogeneity in different regions. This comparative study can help users to understand the advantages and limitations of each fitting method, and choose the appropriate method to reduce the errors and improve the accuracy in their applications.