A Streamlined Solution to Efficiently Reduce Errors and Label Geographical Images

Authors

  • Abishek Kanthan Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA
  • Theodore Spanbauer Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA
  • Chaowei Yang Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA

Abstract

Given the recent rise in global temperatures, the impact of global warming is becoming increasingly significant. To gain a better understanding of global warming’s effects, scientists analyze images that show changes over time with climate – such as Arctic ice. At the status quo, there is a lack of Arctic ice training data, and analyzing images is a tedious and time-consuming task, requiring scientists to manually segment and label their images. To solve this issue, Class X was developed to provide an alternative solution that could complete the same tasks in a simpler and time-efficient manner. Class X uses Python and depends heavily on initial data to train off of as well as manually segmented and labeled data, allowing Class X to autonomously segment and label images. The machine learning model uses datasets such as NASA’s IceBridge data portal to train the machine learning model. As a result, the image classification model for Class X uses aerial Arctic ice images to train along with object-based analysis (OBIA) methods and high spatial resolution data. Additionally, a crucial step in ensuring the software works is creating training data testing model results. As a result, each amenity of Class X enables the software to be a vital tool for scientists around the globe. Class X helps save time and money for companies and allows for data to be easily dissected and used for further studies.

Published

2024-10-13

Issue

Section

College of Science: Department of Geography and Geoinformation Science