An object-based hierarchical classification method for nature reserve land cover classification

Abstract

This study presents an object-based hierarchical classification method for nature reserve land cover classification using hyperspectral and multi-spectral data. The method firstly extracts several indices to identify non-vegetation land covers that are distinguishable with these indices, and then classify vegetation into grass land and crop. The classified land covers were finally assigned to image objects. In this study we obtained an overall classification accuracy of 95.05, with a Kappa of 0.89, which indicates the potential of this method in nature reserve change monitoring and management.