Classification of regional land cover in ALOS PALSAR’s FBD data based on support vector machines

Hongfu Wang and Xiaorong Xue

Abstract

This paper presents the assessment of ALOS PALSAR orthorectified FBD data for regional land cover classification. The SVM approach is mainly used throughout this paper. It is shown that the SVM-RFE algorithm is effective for providing an optimized set of textural parameters to be computed at large scale. In addition to this, an original methodology has been implemented with the intention to give a real insight about the usefulness of textural parameters within the SVM based classification. An optimization of the independent SVM-based classifiers and a clustering procedure complete the methodology.