作　　者： ; ; ; (蒋雪冰）; (谭正喜）;

机构地区： 华南农业大学资源环境学院

出　　处： 《科技通报》 2017年第2期137-142,共6页

摘　　要： 由于归一化植被指数(NDVI)时间序列数据含有大量的噪声,对其应用带来诸多不便。为了提高NDVI数据质量,本文采用线性内插的扩展卡尔曼滤波(EKF)法对广州市森林地区的NDVI时间序列数据进行了重构,并与EKF和中值滤波方法进行比较。利用部分样点的实测数据与重构后的NDVI值进行比较,得到基于线性内插的EKF、EKF和中值滤波三种方法的相对误差分别在-1.91%~0.93%,-3.86%~5.85%和-0.28%~16.30%之间。结果表明:基于线性内插的EKF算法的时间序列重构方法重构后的NDVI时间序列能够更好的逼近高质量的数据,拟合原始曲线的波峰,在提升曲线的整体效果的同时,降低原始数据的均值偏差和数据的离散程度,对低值噪声的抑制能力更好。通过该重构方法重构后的较高质量的NDVI时间序列数据为森林监测、生态保护以及建设提供了良好的基础。 NDVI time-series data contain disturbances that limit their use. In order to improve the quality of NDVI data, this paper combined Linear- Interpolation with Extended Kalman Filter method to reconstruct NDVI time-series data. This study was conducted in forest areas area of Guangzhou city. The accuracies of the predictions from both EKF and Median filter were compared based on the field observations collected from sample plots. Compared to the observed and simulated NDVI values, the relative errors from the method combined Linear-Interpolation with Extended Kalman Filter, Extended Kalman Filter and Median filter varied from - 1.91% to 0.93% , from - 3.86% to 5.85% and, from - 0.28% to 16.30% respectively. The results showed that the method combined Linear-Interpolation with Extended Kalman Filter can be better approximation of high-quality data, fitting the peaks of original curve, which enhances the overall effect of the curve whilst decreasing the mean deviation of the original data’and restraining low noise better. This study implied that this combined Linear-Interpolation with Extended Kalman Filter method increased the reconstructed accuracy of high-quality long time-series NDVI, which provide more high quality NDVI for monitoring forest, ecological protection and construction.

关 键 词： 线性内插 时间序列 扩展卡尔曼滤波 重构

领　　域： [自动化与计算机技术] [自动化与计算机技术]