Landfill Net: A Convolutional Neural Network (CNN) Architecture for the Detection of Landfills from Satellite Imagery in the Continental United States

This project presents a Convolutional Neural Network (CNN) model that identifies landfills in the U.S. using image data, achieving a 97.1% test accuracy. By supporting the creation of a national landfill database, the model can help detect illegal sites and guide the expansion of methane capture systems to reduce greenhouse gas emissions.

As we continue through the 21st century, the effects of climate change have become more and more apparent, with methane emissions playing a pivotal role as one of the most potent greenhouse gasses. Landfill decomposition is a large contributor to this rise in methane, and certain landfills have even adopted methane capture systems that enable the conversion of this methane into valuable energy resources. To continue to mitigate these emissions, it is imperative to establish a comprehensive United States landfill database. We present a Convolutional Neural Network (CNN) model that uses image data in the United States to distinguish whether an image is a landfill or not. The model had a test accuracy of 97.1%, a precision of 96.3%, a recall of 98.1%, and an F-1 score of 97.2%. The results show that the model holds the potential to uncover illegal landfills within the United States and facilitate an informed expansion of methane collection systems in landfill sites.

Explore More!

Research Paper

Source Code