Unmanned Aerial System (UAS) Assisted Framework for the Selection of High Yielding Cultivars

Mounted with advanced sensors onboard, UAS enables the acquisition of crop data at spatial and temporal scales previously unobtainable via traditional remote sensing methods. A novel framework to select cotton genotypes based on data acquired from UAS is developed in this study. Fine resolution (sub-centimeter) orthomosaic images are generated from images acquired from UAS five days before harvesting. Structure from Motion (SfM) algorithm is used in this process. Individual open boll count is then extracted from the orthomosaic image, and structural characteristics such as diameter, perimeter, and area are estimated by applying a series of image processing algorithms. The structural characteristics of cotton bolls are summarized by plot, and the summary statistics of 48 varieties (with 3 replications) are used to select superior genotypes within a breeding population. Our experimental results showed that 75% of high-yield genotypes (based on actual harvest data) can be successfully selected and final yield of the population can be improved by 10% by applying the proposed framework to select cotton genotypes.