Waterlogging Detection in Champaign County with Remote Sensing Imagery and Decision Tree Learning

Wang, Yichen

Permalink

https://hdl.handle.net/2142/106025 Copy

Description

Title

Waterlogging Detection in Champaign County with Remote Sensing Imagery and Decision Tree Learning

Author(s)

Wang, Yichen

Contributor(s)

Schwing, Alexander

Issue Date

2019-12

Keyword(s)

• waterlogging detection
• remote sensing
• decision tree

Abstract

Flooding has become the leading unresolved factor for maize yield loss. Extreme rainfall can cause large and separate waterlogging areas in crop fields, which makes loss prediction difficult. Current waterlogging detection projects usually apply traditional statistical models on public satellite imagery or drone imagery, which usually overlook the lack of resolution or scalability. In this research, we will solve these problems with high-resolution and wide-availability satellite imagery and decision tree learning models. 3-meter resolution PlanetScope CubeSat imagery is used in this research project. As no labels attached to this dataset, our team hand-labeled over two hundred satellite images and converted them to pixel labels. Then, decision tree models and random forest models are trained using these labels. We apply trained models to create pixel-by-pixel waterlogging maps in 2019 Champaign County, and finally, achieved above 94% accuracy.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/106025

Owning Collections