Faculty Mentor(s)

Dr. Darin Stephenson, Mathematics and Statistics; Dr. Brian Yurk, Mathematics and Statistics

Document Type


Event Date



The Hope College Dune Group has been studying West Michigan sand dunes for over twenty years. The group’s interests include observing the mechanisms and effects of sand transport, as well as learning how sand movement and resident dune vegetation affect one another. One of the fundamental tasks of this group is to use machine learning algorithms to create accurate ground-surface and vegetation models from drone imagery in an automated way. A key step in this process is the identification of various types of surface coverage–such as sand, live grass, trees, and other vegetation–automatically from images. An eventual goal of this work is automatic land cover classification at the complex-wide scale.

The scale of the images ranges from high-resolution photos taken with digital cameras to orthomosaics of entire dune complexes taken remotely from a height of around 120 meters. This gives rise to the need for automated alignment and accurate coregistration of multiple images. One technique for image alignment involves using artificial neural networks to identify key points in two or more images and match sets of key points between images. In this poster, we will report on our work on land type classification using a variety of image classification techniques. We produce detailed classifications of high-resolution, low altitude images and use this as a template for creating similar classifications from high-altitude, lower-resolution imagery. We also report on early attempts to align images from multiple perspectives and sources. If automatic image alignment is successful, multiple images can be used together in network training and prediction, and some field-based data collection workflows can be streamlined.


Research reported in this publication was supported in part by funding provided by the National Aeronautics and Space Administration (NASA), under award number 80NSSC20M0124, Michigan Space Grant Consortium and by the Jay Folkert & Charles Steketee Mathematics Summer Research Fund.

Included in

Mathematics Commons