Change Detection and Image Time-Series Analysis 1

2Change Detection in Time Series of Polarimetric SAR Images

Knut CONRADSEN¹, Henning SKRIVER¹, Morton J. CANTY² and Allan A. NIELSEN¹

¹Technical University of Denmark, Kongens Lyngby, Denmark

²Formerly Jülich Research Center, Germany

2.1. Introduction

In this chapter, we will consider the change detection problem in a time series of polarimetric SAR (Synthetic Aperture Radar) images using the covariance representation of multilook polarimetric SAR data. The pixels will then be represented by the complex Wishart distributed Hermitian matrices. The change detection pipeline consists of an omnibus test for testing equality over the whole time span and a subsequent factorization used in assessing individual change time points. The method is easily extended to change detection of “homogeneous” areas, and finally, we will introduce a concept for directional change using the Loewner ordering. The methods are illustrated using airborne as well as satellite data. References to relevant software are provided.

In the first section, we briefly give the background for using an omnibus test instead of pairwise analyses. A more rigorous exposition is given in (Conradsen et al. 2016). Furthermore, we introduce some basic concepts from statistical testing theory, and – in the second section – illustrate their usage of in simulation studies in the gamma distribution.

Figure 2.1. An idealized ...