An Algorithm for Architectural Modeling from a Point Cloud

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Title: "An Algorithm for Architectural Modeling from a Point Cloud"

by

Mr. XU Yisheng

Abstract

As building modeling has broad applications in 3D map, virtual reality and
other aspects, for instance, Google Earth, Virtual Tour and online city
maps all need related technics. Algorithms to extract buildings from point
cloud become important part of these projects.

Fischler and Bolles introduced RANdom SAmple Consensus method to estimate
parameters of a mathematical model from a set of observed data which
contains outliers. This method can be used to extract line and plane from
point data.

In this work, we will introduce the RANSAC to the building modeling. we
first eliminate most points which do not belong to the building, so the
scale of point data is reduced. Then we applied the algorithm based on
RANdom SAmple Consensus method to the modified point cloud. By this
algorithm, we can extract most planar faces of the building. Our
experiments have show the effectiveness and efficiency of the algorithm.


Date            :       27 Aug 2010 (Friday)

Time            :       11:00am to 11:40am

Venue           :       3402 (lift 17-18)

Advisor         :	Professor Long Quan

2nd reader      :	Professor Huamin Qu