A Machine Learning and Deep Learning Approach for Accurate Crop-Type Mapping Using Sentinel-1 Satellite Data

Abstract

Crop classification offers relevant data for crop management, ensuring food safety, and developing agricultural policies. Mapping the crops with high resolution has great significance in determining the position of the crop and effective agricultural monitoring. However, high data costs and poor temporal resolution of satellite data make it difficult to detect the different crops in the field. Therefore, the goal of this survey is to provide an effective analysis of various land cover maps in agriculture using a time series of C-band Sentinel-1 synthetic aperture radar (SAR) data. The various methods based on vertical transmit-vertical receive (VV) and vertical transmit—horizontal receive (VH) polarizations are analyzed to produce better accuracy value for different types of agricultural land. This survey analyzed the different types of existing classification methods such as machine learning and deep learning algorithm used in Sentinel-1 satellite data. The overall accuracy (OA), kappa coefficient, user accuracy (UA), producer accuracy (PA), and F1-score are considered key parameters for defining the effectiveness of crop-type classification in land cover types. This comprehensive research supports the researchers to obtain the best solutions for the current issues in crop-type mapping using Sentinel-1 SAR data.