Features extraction from multi-date ASTER imagery using a hybrid classification method for land cover transformations

This work proposes a features extraction strategy for each land cover class using a hybrid classification method on multidate ASTER data. To enable an effective comparison among multi-date images, Multivariate Alteration Detection (MAD) transformation was applied for data homogenization to reduce noises due to local atmospheric conditions and sensor characteristics. Consequently, different features identification procedures, both spectral and object-based, were implemented to overcome problems of misclassification among classes with similar spectral response. Lastly, a postclassification comparison was performed on multi-date ASTER-derived land cover (LC) maps to evaluate the effects of change in the study area. All the above methods, when used in multi-date analysis, do not consider the issue of data homogenization in change detection to reduce noises due to local atmospheric conditions and sensor characteristics.