Generating Digital Terrain Models From Aerial LIDAR Using Convolutional Neural Networks
Open Access
Author:
Pugh, Brian
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Manuel Gonzalez Rivero, Thesis Supervisor Jeffrey Scott Mayer, Thesis Honors Advisor
Keywords:
convolutional neural network digital elevation terrain ground filtering
Abstract:
Presently, airborne LIDAR data measures the elevation of the highest point observed rather than the underlying bare-earth elevation. Digital Terrain Model generation from LIDAR poses an important problem in remote sensing. Many solutions rely on labeling and removing non-ground data points and interpolating the unmeasured ground from the surrounding ground points. In this paper we show that a feed-forward convolutional neural network can be successfully trained and used to directly discover the underlying terrain model from collected LIDAR data.
