Date of Graduation

Spring 2019

Degree

Master of Natural and Applied Science in Geography, Geology, and Planning

Department

Geography, Geology, and Planning

Committee Chair

Xin Miao

Keywords

image segmentation, watershed, tree extraction, reconstruct 3D model, unmanned aerial system, UAS, DSM

Subject Categories

Geographic Information Sciences | Programming Languages and Compilers | Spatial Science

Abstract

This project aims to develop and assess methodology for spatial modeling and extracting individual trees from high spatial resolution Digital Surface Model (DSMs) derived from unmanned aerial system (UAS) or drone-based aerial photos. Those results could be used for monitoring of vegetative response of forests, grasslands and vineyards to regional and localized fluctuations in climate and seasonality. The primary objective of this research is to extract 3D spatial information using drone-based aerial imagery through photogrammetric methods. UAS flights were taken place at phenologically critical times over several locations owned and managed by Missouri State University (MSU). The 3D DSM can be reconstructed from obtained images through photogrammetric software. And then a computer programming language (Python) is used to develop an algorithm to extract every single tree from 3D models. First, DSM data is imported into Python as a raster data layer. Next, the watershed segmentation algorithm with two different image filtering is used for single tree extraction, and the accuracy comparison is conducted between the two methods. The number of the trees can be used to calculate vegetative metrics and growth rate if multi-year data is available in the future. And time-series data can be correlated with climate fluctuation and seasonality. This study can provide a new approach for monitoring various types of natural and agricultural vegetation to aid in making short-term and long-term management decisions.

Copyright

© Hai Ha Duong

Available for download on Tuesday, May 17, 2022

Open Access

Share

COinS