知识引导的稀疏时间序列遥感数据拟合

发布时间：2018-08-12 19:01

【摘要】：在多云多雨的地区,光学遥感存在着获取无云数据困难的难题,这会导致时间序列应用中可用数据匮乏。因此,本文面向稀疏时间序列遥感数据,根据噪声造成遥感影像上归一化差分植被指数(NDVI)被低估的事实,提出了一种知识引导的拟合方法。首先,在遥感影像预处理的基础上,利用先验知识和时序差分法对噪声进行识别和剔除;然后,采用高斯二阶模型对原始数据进行拟合;最后,根据拟合残差更新权重,进行迭代拟合,重复上述过程直至获得稳定的结果。本文以Landsat 8 OLI作为数据源,对浙江省杭州地区的森林数据进行拟合,结果表明:在稀疏时间序列数据的情况下,本文方法与MODIS数据拟合结果的相关系数达到0.92,关键时点(如NDVI峰值点等)的时间误差在5 d;相比当前主流方法的0.88与8 d具有更高的精度。
[Abstract]:In cloudy and rainy areas, optical remote sensing is difficult to obtain cloudless data, which leads to the lack of available data in time series applications. Therefore, in this paper, for sparse time series remote sensing data, according to the fact that the normalized differential vegetation index (NDVI) is underestimated in remote sensing images caused by noise, a knowledge guided fitting method is proposed. First, on the basis of remote sensing image preprocessing, the prior knowledge and time-series difference method are used to identify and eliminate the noise. Then, the Gao Si second-order model is used to fit the original data. Finally, the weight is updated according to the fitting residual error. Iterative fitting is carried out and the process is repeated until a stable result is obtained. In this paper, Landsat 8 OLI is used as the data source to fit the forest data in Hangzhou area of Zhejiang Province. The results show that in the case of sparse time series data, The correlation coefficient between this method and the fitting results of MODIS data is 0.92.The time error of critical time points (such as NDVI peak) is 5 days, which is higher than that of the current mainstream methods 0.88 and 8 days.
【作者单位】：浙江工业大学计算机科学与技术学院;
【基金】：国家自然科学基金(编号:61572437,41301473)~~
【分类号】：TP751

【相似文献】