Agricultural Fields Monitoring with Multi-Temporal Polarimetric SAR (MT-POLSAR) Change Detection

Abstract

This work presents a novel methodology to extract and analyse multi-temporal polarimetric SAR (PolSAR) information from a stack of co-registered images. The method is based on the analysis of PolSAR changes between every image with respect to the rest of images in the stack. The changes are organized in a matrix form to encode the polarimetric evolution of a target. The change matrix is then used to visually understand a target evolution and its SAR response based on the evolution of scattering mechanisms due to the target physical evolution. Additionally, we design and test an image classification algorithm in a supervised learning fashion by using typical change matrices as training data. The methodology is tested exploiting C-band quad-pol RADARSAT-2 data with special interest on agricultural fields such as rice in Seville, South-West of Spain and in the Indian Head in Canada as part of the Agrisar 2009 campaign.