Copy the page URI to the clipboard
Marino, Armando and Hajnsek, Irena
(2014).
DOI: https://doi.org/10.1109/TGRS.2013.2284510
Abstract
The possibility to detect changes in land cover with remote sensing is particularly valuable considering the current availability of long time series of data. SAR can play an important role in this context, since it can acquire complete time series without limitations of cloud cover. Additionally, polarimetry has the potential to improve significantly the detection capability allowing the discrimination between different polarimetric targets. This paper is focused on developing two new methodologies for testing the stability of observed targets (i.e. Equi-Scattering Mechanisms hypothesis) and change detection. Both the algorithms adopt a Lagrange optimization, which can be performed with two eigen-problems. Interestingly, the two optimizations share the same eigenvectors. Three statistical tests are proposed to set the threshold for the change detector. Two of them are mostly aimed at point targets and one is more suited for distributed targets.
All the algorithms and procedures developed in this paper are tested on two different quad-polarimetric dataset acquired by the E-SAR DLR system in L-band (SARTOM 2006 and AGRISAR 2006 campaigns). The dataset are accompanied by ground surveys. The detectors are able to identify targets and areas with validated changes or showing clear differences in the images. The theoretical pdf exploited to model the optimum ratio fits adequately the data and therefore has been used for the statistical tests. Regarding the output of the tests, two of them provided good results, while one needs more care and adjustments.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)
