Artificial Neural Network Based Cotton Yield Estimation Using Unmanned Aerial System Data

Harvest yield is one of the most critical traits in cotton breeding program and it is also the most time-consuming, labor-intensive, and expensive component in the field experiment. For this reason, field experiment size of breeding trials is often limited by resources available to harvest cotton in the field at the end of the growing season. Having ability to estimate cotton harvest yield without harvesting will open up great opportunities for cotton breeding program to scale up field experiment size significantly, which can result in higher chance to speed up high yielding cultivar development process. The objective of this study is to an artificial neural network algorithm for estimating cotton harvest yield using multi-source and spatio-temporal remote sensing data collected from Unmanned Aerial System (UAS). The proposed algorithm takes UAS based multi-temporal features such as canopy cover, canopy height, canopy volume, Normalized Difference Vegetation Index (NDVI), Excessive Greenness Index along with non-temporal features such as cotton boll count, boll size and boll volume as input and predicts the corresponding yield. UAS data collected in Corpus Christi and College Station, TX in 2016, 2017, and 2018 will be used to train and validate the machine learning model, and cotton yield prediction accuracy will be discussed in the presentation.