Reconstruction of Tree Model with Airborne LiDAR Data and Multi-spectral Images-An Example of NCKU Campus

碩士 === 國立成功大學 === 測量及空間資訊學系碩博士班 === 96 === The vegetation monitoring system, which observes the distribution and transition of trees, is important to ecosystem management. Tree modeling is one of the essential work in forest monitroing. It was expensive and time consuming in the past when ground inv...

Full description

Bibliographic Details
Main Authors: Jin-yi Lai, 賴君怡
Other Authors: Yi-hsing Tseng
Format: Others
Language:zh-TW
Published: 2008
Online Access:http://ndltd.ncl.edu.tw/handle/65426402585969599668
Description
Summary:碩士 === 國立成功大學 === 測量及空間資訊學系碩博士班 === 96 === The vegetation monitoring system, which observes the distribution and transition of trees, is important to ecosystem management. Tree modeling is one of the essential work in forest monitroing. It was expensive and time consuming in the past when ground investigation was the major means of tree modeling. The purpose of this study is to build tree models using remote sensing technologies. The objective of this study is to build tree models using LiDAR data and aerial multispectral images. The proposed scheme comprises three major steps: (1) tree cloud points extraction; (2) tree crown detection; (3) tree modeling. Tree cloud points extraction is a filtering algorithm to exclude non-tree cloud points using DEM and multispectral images. In the tree crown detection, watershed segmentation is used to find the crown delineation, then it searches the local maxima of the segments and performs roundness test to obtain the tree parameters. Finally, parameters of tree models are adopted for tree modeling. The test data for this study include digital aerial multispectral images and LiDAR data of National Cheng Kung University. Modeling of some banyan trees in the campus is demonstrated. The test shows that indivisual trees can be modeled well. However, watershed segmentation frequently failed when trees are closely located. In this case, the roundness test can obtain better parameters for tree modeling.