Recursive classification of satellite imaging time-series: An application to land cover mapping

A wide variety of applications of fundamental importance for security, environmental protection and urban development need access to accurate land cover monitoring, for which the analysis of optical remote sensing imagery is key. Classification of time-series images, particularly with recursive methods, is of increasing interest in the current literature. Nevertheless, existing recursive approaches typically require large amounts of training data. This paper introduces a recursive classification framework that improves the decision-making process in multitemporal and multispectral land cover classification algorithms while requiring low computational cost and minimal supervision. The proposed approach allows the conversion of an instantaneous classifier into a recursive Bayesian classifier by using a probabilistic framework that is robust to non-informative image variations. Three two-class experiments are conducted using Sentinel-2 data. The first one consists in the water mapping of an embankment dam in Oroville, California, USA; the second one is a water mapping experiment of the Charles river basin area in Boston, Massachusetts, USA; and the last experiment addresses deforestation detection in the Amazon rainforest. The spectral index classifier (SIC) is introduced as a method to convert broadband spectral indices such as the NDWI and the MNDWI into probabilistic classification results. A classifier based on the Gaussian mixture model (GMM), a logistic regression (LR) classifier, and the SIC are compared to their recursive counterparts. Results show that the use of Bayesian recursion significantly increases the robustness of existing instantaneous classifiers in multitemporal settings, including two state-of-the-art deep learning-based models, without the need for additional training data. Furthermore, the proposed framework scales well for an increasing number of classes.