Land cover classification of the Alps from InSAR temporal coherence matrices

Land cover mapping is of great interest in the Alps region for monitoring surface occupation changes (e.g. forestation, urbanization, etc). In this pilot study, we investigate how time series radar satellite imaging (C-band single-polarized SENTINEL-1 Synthetic Aperture Radar, SAR), also acquired through clouds, could be an alternative to optical land segmentation. Concretely, compute every location (using SAR pixels over 45 × m ) temporal coherence matrix Interferometric (InSAR) phase 1 year. This normalized size 60, ×, 60 (60 acquisition dates year) summarizes reflectivity land. Two machine learning models, a Support Vector Machine (SVM) and Convolutional Neural Network (CNN) have been developed estimate classification performances 6 main classes (such as forests, urban areas, water bodies, or pastures). The training database was created by projecting geometry reference labeled CORINE Cover (CLC) map on mountainous area Grenoble, France. Upon evaluation, both models demonstrated good with overall accuracy 78% 81% Chambéry (France). We show how, even spatially coarse database, our model able generalize well, large part misclassifications are due low precision ground truth map. Although some less computationally expensive approaches data) available, based very different information, i.e., patterns year, complementary thus beneficial; especially regions where not always available clouds. Moreover, that InSAR informative, which lead future other applications such automatic detection abrupt snow fall landslides.