遥感时间序列数据滤波重建算法发展综述  

Review on methods of remote sensing time-series data reconstruction

作　　者：李儒[1,2] 张霞[1] 刘波[1,2] 张兵[3] 

机构地区：[1]中国科学院遥感应用研究所遥感科学国家重点实验室,北京100101 [2]中国科学院研究生院,北京100039 [3]中国科学院对地观测与数字地球科学中心,北京100080

出　　处：《遥感学报》2009年第2期335-341,共7页NATIONAL REMOTE SENSING BULLETIN

基　　金：国家自然科学基金项目(编号:40601069);遥感科学国家重点实验室开放基金(编号:03Q00300649)

摘　　要：遥感时间序列数据(MODIS,NOAA/AVHRR,SPOT/VEGETATION等)在植被生长监测、物候信息提取、土地利用类型监测等诸多领域得到了广泛应用,是生产研究的重要数据源之一。由于传感器、云层大气等影响,遥感时间序列数据存在着严重的噪声,应用前必须进行序列滤波重建工作。综述现有各类滤波重建方法,对研究中广为采用的3类主要方法(基于最小二乘的非对称高斯函数拟合、Savitzky-Golay滤波、基于离散傅里叶的系列分析方法)集中阐述其理论基础、应用步骤和优缺点。总结当前遥感时间序列滤波重建方法需要进一步改进之处。Remote sensing time-series have been applied successfully in various fields, such as vegetation change monitoring, phonological （seasonality） information extraction, land use dynamic classification etc. It is one of the most important data sources for kinds of research work ＆ engineering project. However, due to the effect of sensor, cloud, and atmospheric conditions, there are serious residual noise in time-series data. Therefore, prior to further applications, it is need to filter residual noise to reconstruct the series. Many methods have been developed to solve this problem. In this paper, the following methods are summarized first, including Maximum Value Composite （MVC）, Best Index Slope Extraction Algorithm （BISE）, Temporal Windows Operation （TWO）, Asymmetric Gaussian Function Fitting Approach （AGFF）, Savitzky-Golay Filtering （S-GF）, Harmonic Analysis Algorithm （HAA）, Local Maximum Fitting （LMF）. MVC is more acceptable than other methods because it is useful when producing remote sensing time-series products. But the products are the primary ones and contain much residual noise. This method is helpless to reconstruct the series further. In fact, the data needed to reconstruct before applications of the products made according to MVC. BISE uses a sliding time window to capture local maxima. It requires the determination of the sliding period and a threshold.for acceptable percentage increase. TWO can reduce the noise caused by cloud and atmosphere without auxiliary data. However, the requirement of the parameters from experience limits its application. AGFF and S-GF are two strategies developed in recent years. LMF, compared with HAA, first filt.ers noise and then reconstructs the data processed. Then three most frequently-used approaches, AGFF, S-GF and HAA, are introduced in detail in terms of the basic theories, application steps and advantages ＆ disadvantages. AGFF employs more than two combinative Gauss-shape curves to fit the series. Every combination simulate

关 键 词：遥感时间序列 滤波 重建 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]