Cloud Implementation of Country Level Crop Classification Based on Time Series of Satellite Data

Crop classification and mapping is a cordial task for many remote sensing applications. Nowadays many countries are moving to operational agricultural monitoring which requires early season crop classification and mapping as well as assessment of crop state and conditions. For this problem, freely available optical and SAR data from Sentinel-1 -2 as well as Landsat-8 are very helpful, because they allow using dense time series as input information for classification with different machine learning algorithms [1]. But the opportunities are also lead us to the problem of Big Data, especially for countries with large territories which should be solved with high performance computational techniques [2]. To solve the problem we have implemented satellite data processing and classification algorithms in cloud environment of Amazon Web Services. The main advantages of such implementation are as follows: (i) fast access to the data, which are stored at the same location and avoiding data downloading stage; (ii) scalability of the classification process, when we classify different territories simultaneously with as many instances as we need. Within such an approach we have implemented different classification algorithms, as our own Deep Learning Neural Network classifier [3] as well as open source Sen2Agri system [4], and use the system for crop classification in Ukraine within World Bank project and other countries of Eurasia for decision making support.

1. Shelestov Andrii, et al. "Exploring Google Earth Engine Platform for Big Data Processing: Classification of Multi-Temporal Satellite Imagery for Crop Mapping." Frontiers in Earth Science, 5.17, pp.1-10, 2017. 2. Shelestov, A.Y., Kussul, N.N. "Using the fuzzy-ellipsoid method for robust estimation of the state of a grid system node." Cybernetics and Systems Analysis, 44.6, pp 847-854, 2008. 3. Kussul, Nataliia, et al. "Deep learning classification of land cover and crop types using remote sensing data." IEEE Geoscience and Remote Sensing Letters, 14.5, pp. 778-782, 2017. 4. Kussul, Nataliia, et al. "Sentinel-2 for agriculture national demonstration in Ukraine: results and further steps." IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 5842-5845, 2017.