In order to join virtual sessions, you must be registered and logged-in(Were you registered for the in-person meeting in Denver? if yes, just log in.) 
Note: All session times are in Mountain Daylight Time.

Lidar-Based Machine-Learning Model to Identify Archaeological and Natural Features in the Maya Lowlands of Belize, Guatemala, and Mexico

Authors: Leila Character*, University of Texas - Austin, Tim Beach, University of Texas at Austin, Cody Schank, University of Texas at Austin
Topics: Spatial Analysis & Modeling, Geomorphology
Keywords: Machine-learning, Deep-learning, AI, Geomorphology, Geomorphometry, Archaeology, Caves
Session Type: Paper
Presentation File: No File Uploaded


The authors are developing a lidar-based computer model that uses convolutional neural networks to identify ancient Maya landscape features under dense Central American forest canopy. Thus far, we have completed and ground-verified a model that uses a random forest classifier to identify cave entrances at a 260-km2 Maya site in Belize. We are now expanding this work to generate a predictive model for identification of undiscovered, overgrown ancient Maya architecture in Guatemala and Mexico. Recent applications of sub-meter resolution lidar imagery to such sites has revealed huge areas of previously unknown architecture. However, it has all been identified by slow, hit-or-miss visual inspection. We are automating this process by generating a machine-learning model based on a large training dataset, drawing from the fields of computer vision and machine-learning. This work will help archaeologists and cave scientists to quickly identify study areas and produce accurate maps for efficient management of resources. Additionally, models that can automatically identify features of interest using aerial imagery are highly relevant across a broad range of fields both inside and outside academia, including natural hazard prediction and urban planning.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login