Sentinel-2 Multidimensional data cubes for crop monitoring time series classification

Geoprocessing and remote sensing play an important role when it comes to monitoring land use and land cover using large volumes of data (Big data). In this context, Satellite Time Series Image (Data Cubes) emerge as an alternative to manage Big data mining and classification. Combining information and describing data using time series analysis methods, like Time-Weighted Dynamic Time Warping (TWDTW), for pattern recognition and classification in diverse areas, becomes possible to observe and understand land use and land cover changes as agricultural expansion and crop monitoring. Thus, this work aims to classify crops dynamics in the western portion of Bahia - Brazil, using machine learning and data cubes. Our results showed consistency and feasibility in mapping agricultural targets on a monthly base, with a reasonable classification accuracy over 70% for the produced maps.