2021/11/15 by Giorgio Morales, Morales, Giorgio, John Sheppard +1
Agricultural and Biological Sciences · Environmental Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Remote Sensing in Agriculture #Smart Agriculture and AI #Soil Moisture and Remote Sensing #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2111.08069
openalex publication_date 2021/11/15 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Crop yield prediction is one of the tasks of Precision Agriculture that can\nbe automated based on multi-source periodic observations of the fields. We\ntackle the yield prediction problem using a Convolutional Neural Network (CNN)\ntrained on data that combines radar satellite imagery and on-ground\ninformation. We present a CNN architecture called Hyper3DNetReg that takes in a\nmulti-channel input image and outputs a two-dimensional raster, where each\npixel represents the predicted yield value of the corresponding input pixel. We\nutilize radar data acquired from the Sentinel-1 satellites, while the on-ground\ndata correspond to a set of six raster features: nitrogen rate applied,\nprecipitation, slope, elevation, topographic position index (TPI), and aspect.\nWe use data collected during the early stage of the winter wheat growing season\n(March) to predict yield values during the harvest season (August). We present\nexperiments over four fields of winter wheat and show that our proposed\nmethodology yields better results than five compared methods, including\nmultiple linear regression, an ensemble of feedforward networks using AdaBoost,\na stacked autoencoder, and two other CNN architectures.\n