Multi-Temporal Sentinel-2 Images for Classification Accuracy

Abstract

The term ‘multi-temporal’ refers to data recorded by sensors scanning the same scene at different dates and times. In this research, Sentinel-2A images were selected for the analyses. Sentinel-2 is a very new program of the European Space Agency (ESA) designed for fine spatial resolution global monitoring. Land Cover–Land Use (LCLU) classification tasks can take advantage of the fusion of radar and optical remote sensing data, generally leading to increased mapping accuracy. Here we propose a methodological approach to fuse information from the new European Space Agency Sentinel-1 and Sentinel-2 imagery for accurate land cover mapping of a portion of the South Solok region, West Sumatra. Data pre-processing used the European Space Agency’s Sentinel Application Platform and the SEN2COR toolboxes. The two main objectives of this study are to evaluate the potential use and synergetic effects of ESA Sentinel-1A C-band SAR and Sentinel-2A Optical data for classification and mapping of LCLU. First, the pre-processing chain supported by sensor-specific toolboxes developed by ESA represents a reliable and fast approach for the preparation of ready-to-process imagery. Second, an investigation to derive a methodological framework to integrate Sentinel-1 and Sentinel-2 imagery for land cover mapping by integrating radar and optical imagery has been set up and tested. The results of object-based classification produced higher accuracy than pixel-based classifications if the same type of classifier is used. The classifier algorithms using k-Nearest Neighbors (k-NN) function performed slightly better, with an overall accuracy of 91.30%.