Supervised classification and NDVI calculation from remote piloted aircraft images for coffee plantations applications

Abstract

A tool that has been widely used in Precision Agriculture (PA) is the Remote Piloted Aircraft’s (RPA’s). These tools are used to monitor crops, in addition to checking and quantifying various attributes related to plants. However, there are few studies that evaluate the applicability of this technology in coffee plantations. The objective of this study is to present the applicability of two tools associated with PA and remote sensing to monitoring a coffee planta-tion. The study was conducted in the municipality of Três Pontas, Brazil, comprised a 1.2 ha coffee plantation. Data were collected during a flight with an eBee SQ RPA, and high spatial resolution images were captured by a Parrot Sequoia multispectral sensor coupled to the aircraft. The images were processed using the software Pix4D, thus creating an orthomosaic that was later uploaded to QGIS software. In this program, a supervised classification of land use and land cover was performed using the maximum likelihood method, and the following classes were obtained: coffee plant, exposed soil, and undergrowth. From the mapping accuracy, an overall accuracy and kappa index of 91% and 85% were obtained, respectively. In addition to the supervised classification of the site, the normalized difference vegetation index (NDVI) was calculated for only the coffee plant class. The NDVI map showed the areas of the plantation coffee crop with higher and lower vegetative vigour.