Classification of Cynodon spp. grass cultivars by UAV

Abstract

Traditional methods for estimating biomass in pasture frequently use destructive methods with high demand for time, resources and labor. The development of models for automated estimation of biomass and leaf area index, particularly from images captured by Unmanned Aerial Vehicle (UAV), saves resources and helps the adoption of anticipatory measures in the management of the experimental area. The objective of this study was to create a technical feasibility study for the use of UAV in the estimation of biomass, forage canopy height, and general conditions of Cynodon grass in plots, using volume and vigor by the radiometric and morphometric approach, the NDRE index, and digital terrain (DTMs) and digital surface (DSMs) models compared to scores by the specialist in the field. Visible (RGB), red edge (RedEdge) and near infrared (NIR) imaging cameras were used for continuous monitoring of the experimental area, of approximately 3,800 m2, located at the José Henrique Bruschi Experimental Field (CEJHB), in the municipality of Colonel Pacheco, Minas Gerais, Brazil. After UAV imaging, we selected nine Cynodon spp. clones that showed greater vigor based on the data from the field plots and data obtained by UAV and classified using the method to estimate the vegetation vigor index (VVI) and classified by natural breaks in GIS.