Classification of organic quinoa crops using multispectral aerial imagery and machine learning techniques

Abstract

Crop mapping is a vital tool for agricultural management and food security that can benefit from remote sensing data. The purpose of this research is to use machine learning (ML) techniques to classify quinoa crops from multispectral images. The spectral reflectance of five optical bands is used to develop classification models that are tested for diverse quinoa phenological phases. Deep learning methods Segnet and Unet were investigated, as well as Decision Trees, Discriminant Analysis, K nearest Neighbor, Support Vector Machines, Adaboost and Random Forest. Data was collected from quinoa crop fields in Cabana, Puno region in Peru. The multispectral images were captured using an Unmanned Aircraft System (UAS) from a height of 50 meters. Deep learning methods leave behind other approaches in the classification job, according to the results.