Deep Learning for Classification and Growth Prediction of Turmeric and Sugarcane Crops

The management of agricultural output and the formulation of agricultural policies, which affect the stability of the nation and society, both heavily depend on crop yield. Access to crop spatial distribution in real time is essential because of this. This study gathered information mostly from SENTINEL-2A satellite and developed a system for classifying crops primarily on phenological variables that were compared and analysed, as well as picture segmentation techniques. The profound principle of this method is to first use training data to generate a standard phenological feature sequence (SPFS) for each crop, and then use the compartmentalized test data and the vegetative growth pattern for every crop to determine the kind of surface features by calculating similarity coefficient between two sets of data. The findings indicate that phenological features and feature selection are valuable in enhancing crop generalization performance, with the generalization ability on a cross basis being 76.06%. For agricultural monitoring, multispectral SAR data have significant promise.