像素级SAR影像时间序列的建模方法研究

Research on Modeling Method for Pixel-Level SAR Image Time Series

像素级遥感影像时间序列保留了更多的原始影像信息和地表细节层次信息,能够更准确地提取地物信息、揭示地物的变化规律。针对SAR影像堆积数据,提出了一种像素级SAR影像时间序列的建模方法,主要步骤包括辐射校正和高精度的几何配准,将SAR影像数据集转换为一系列由离散的后向散射系组成的时间序列文本化数据。利用2006-2009年21幅分辨率为150m的WSM模式ENVISAT ASAR影像进行建模试验,构建了像素级SAR影像时间序列。试验表明:该建模方法具有较高的几何精度和辐射精度,能够直接应用于地物识别,在地物信息提取、聚类分析、变化检测等方面发挥独特的优势。

Pixel-level remote sensing image time series retain more raw information and detail level information on surface features. It contributes to extract feature information more accurately and reveal the surface change rules. Aimed at SAR image data set, this paper presented the modeling method of pixel-level SAR image time series, the main procedure of which includes radio- metric correction and high-precision geometric registration. The method transforms SAR images stacked data into time series textual data consisting of discrete backscattering coefficients. 21 wide-swath ENVISAT ASAR images from 2006 to 2009 which have 150 m resolution with a pixel spacing of 75m were utilized to model pixel-level SAR image time series. Geometric accuracy and radiometric accuracy were evaluated respectively in the experiment. The experimental results showed that the modeling method possesses high precision both in geometry and radiation and applied to identify the ground object type. It will play a u- nique advantage in features extraction, clustering analysis and change detection, bearing strong potential in application.