Simultaneously Estimating Crop Yield and Seed Composition using Multitask Learning from UAV Multispectral Data
Document Type
Conference Proceeding
Publication Date
10-20-2023
Department
Department of Applied Computing
Abstract
This study focused on simultaneous estimation of corn grain yield and protein concentration using multitask deep learning. Unmanned Aerial Vehicles (UAV) multispectral data was collected throughout the 2022 growing season from a corn field located near Volga, South Dakota, USA. The multispectral data was used to derive canopy spectral and textural features, which were used as input variables to the deep learning models. Multitask Learning (MTL) approach based on fully connected forward deep neural network (DNN) and one dimensional convolutional neural network (1D-CNN) was used to predict corn grain yield and seed protein concentration simultaneously. The results produced by MTL approach were compared to that of single task learning (STL). Results show that MTL approach based on 1D-CNN model produced better results as compared to the STL approach for both yield and protein concentration prediction.
Publication Title
International Geoscience and Remote Sensing Symposium (IGARSS)
ISBN
9798350320107
Recommended Citation
Khan, S.,
Maimaitijiang, M.,
Millett, B.,
Paheding, S.,
Li, D.,
Caffe, M.,
&
Kovacs, P.
(2023).
Simultaneously Estimating Crop Yield and Seed Composition using Multitask Learning from UAV Multispectral Data.
International Geoscience and Remote Sensing Symposium (IGARSS),
2023-July, 2771-2774.
http://doi.org/10.1109/IGARSS52108.2023.10282273
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p2/362