Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.
Phenotyping Agronomic Traits of Peanuts using UAV-based Hyperspectral Imaging and Deep Learning
Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org
Citation: 2022 ASABE Annual International Meeting 2200814.(doi:10.13031/aim.202200814)
Authors: Kamand Bagherian, Rafael Bidese Puhl, Yin Bao, Qiong Zhang, Alvaro Sanz-Saez, Charles Chen, Phat Dang
Keywords: Peanut, High-Throughput Plant Phenotyping, Agronomic Traits, Deep Learning, Hyperspectral Imaging, Unmanned Aerial Vehicle (UAV).
Abstract. Agronomic traits of peanuts, such as biomass, pod count, and pod yield, are utilized for assessing crop vigor, nutrient use efficiency, and efficacy of agricultural management practices. These phenotypes also help identify superior genotypes in peanut breeding programs. However, direct measurements of these traits are destructive and time-consuming. In this study, we assessed the feasibility of using aerial hyperspectral imaging and machine learning to predict biomass, pod count and yield of peanut plants. For this purpose, two different approaches were evaluated. The first approach employed eighty narrow-band vegetation indices as input features. And four machine learning models including K-nearest neighbors (KNN), support vector regression (SVR), random forest (RF), and a multi-layer perceptron (MLP) regressor were combined to create an ensemble model. The second approach utilized the raw hyperspectral cubes to derive mean and standard deviation of spectral reflectance per band. These features were used to train a deep learning model consisting of 1-dimensional convolution layers followed by a MLP regressor. Predictions obtained using feature learning and deep learning (R2 = 0.45-0.73; sMAPE = 24-51%) outperformed those obtained using feature engineering and conventional machine learning models (R2 = 0.44-0.60, sMAPE = 27-59%). The results revealed that the proposed high-throughput phenotyping technology showed promising potential to screen a large number of genotypes and accelerate peanut breeding programs.
(Download PDF) (Export to EndNotes)