An Attention Model for Remote Sensing Landcover Classification
Open Access
- Author:
- Simmons, Shelton
- Graduate Program:
- Spatial Data Science
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- June 22, 2026
- Committee Members:
- Bing Zhou, Thesis Advisor/Co-Advisor
Fritz Connor Kessler, Committee Member
Anthony Robinson, Program Head/Chair - Keywords:
- remote sensing
land-cover classification
multi-spectral
Landsat
deep learning
vision transformer - Abstract:
- Herein, the design of a transformer used to perform remote sensing landcover image classification is analyzed. The transformer was designed to use the traditional remote sensing indices to perform the classification. A custom dataset was created for this study using Landsat 8 and 9 OLI imagery. The dataset consists of 50 images of each landcover class; urban, water, vegetation, agricultural, forest, and bare earth with an additional 10 images per class for validation. Each image was classified using the remote sensing indices to produce square clips that are roughly 90 pixels per side. The vision transformer is trained using this custom training set by giving a bonus value to pixels which fall into the target range for the various remote sensing indices and a penalty for those that are classified while outside of the target range. This transformer had an overall accuracy gain of 1% compared to the transformer that didn’t use thresholds. Furthermore, the reward and penalty logic was shown to reduce the training loss of the transformer.
Accessible Version in Progress
We're generating an accessible version of this file to meet ADA Title II requirements. This process may take up to one hour. Please return later to access the accessible copy once it's ready.
You can still download the current version by clicking "OK".
What's happening:
An accessible PDF is being generated using Adobe with AI used to generate alternative text (alt text) for images in the PDF.