Crop Type Classification by DESIS Hyperspectral Imagery and Machine Learning Algorithms

Developments in space-based hyperspectral sensors, advanced remote sensing, and machine learning can help crop yield measurement, modelling, prediction, monitoring for loss prevention global food security. However, precise continuous spectral signatures, important large-area growth early prediction of production with cutting-edge algorithms, be only provided via imaging. Therefore, this article used new-generation Deutsches Zentrum für Luft- und Raumfahrt Earth Sensing Imaging Spectrometer (DESIS) images to classify the main types (hybrid corn, soybean, sunflower, winter wheat) Mezőhegyes (southeastern Hungary). A Wavelet-attention convolutional neural network (WA-CNN), random forest support vector (SVM) algorithms were utilized automatically map crops over agricultural lands. The best accuracy was achieved WA-CNN, a feature-based deep algorithm combination two overall (OA) value 97.89% user's producer's from 97% 99%. To obtain this, first, factor analysis introduced decrease size image data cube. wavelet transform applied extract features combined attention mechanism CNN gain higher mapping types. Followed by SVM reported OA 87.79%, accuracies its classes ranging 79.62% 96.48% 79.63% 95.73%, respectively. These results demonstrate potentiality DESIS observe different predict harvest volume, which is crucial farmers, smallholders, decision-makers.