Optimized unsupervised CORINE Land Cover mapping using linear spectral mixture analysis and object-based image analysis

In this paper, the approach Linear Spectral Mixture Analysis and Object-Based Image Analysis (LSMA + OBIA) and Iterative Self-Organizing Data Analysis Technique and Object-Based Image Analysis (ISODATA + OBIA) are evaluated, for optimizing land cover mapping in high mountain areas from Landsat-8 multispectral images. Both approaches are applied to generate in a semiautomatic and unsupervised way a land cover map of the Santurban-Berlin moorland, located in Colombia as a case study to carry out the evaluation. It has been found that LSMA + OBIA allows the generation of a land cover classification with a maximum global reliability of 88% compared to a reliability of 79% with ISODATA + OBIA.