利用黑龙江省2012年16 d 250 m的MODIS NDVI数据，使用Savitsky－Golay滤波法、Logistic调合函数模型和非对称Gaussian模型3种方法对曲线拟合重构，引入均方根误差比较3种方法的优劣；在曲线平滑的基础上提取物候参数，分别对森林和农用地用固定阈值和动态阈值提取其物候参数，目的在于比较两种阈值设定方法的优劣，并获取阈值使物候参数结果与作物农事历信息吻合。结果表明， Savitsky－Golay滤波法在与原始曲线近似度方面优于另外两种拟合方法。%This assay is based on the sixteen-day, 250 meters MODIS NDVI dataset in the year 2012, and use three methods-the Savitsky-Golay filter, the Logistic filter and the Gaussian model-respectively to construct the NDVI time series, then use the root-mean-square error to compare the quality of the three methods.Extracting phonological parameters on the base of smoothed curves through setting different dynamical and stationary thresholds to contrast which setting method is better, and to get a appropriate thresh-old that math the seasonal calendar accurately.The results indicate that, the Savitsky-Golay filter is better than the others in respect of the extend of approximating to the crude curves;using the stationary method to extract phonology parameters can match the season calendar more accurately than using stationary method.
展开▼
