Linear Structure Vectorization in Large-Scale Landscape Point Cloud

MPhil Thesis Defence


Title: "Linear Structure Vectorization in Large-Scale Landscape Point Cloud"

By

Mr. Haoan FENG


Abstract:

The advancement of modern remote sensing technology reduces the cost of 
acquiring three-dimensional information in the real-world. The information 
obtained is generally represented in discrete data points known as the point 
cloud. Many researchers focus on algorithms and technologies to extract the 
underlying topology supporting many interesting applications such as indoor 
navigation and heritage reconstruction. In this thesis, we focus on the linear 
and thin structures within the point cloud instead of surfaces and we propose 
an automatic pipeline taking raw point clouds of large-scale landscapes, which 
are generated by the multi-perspective image reconstruction algorithms. Our 
pipeline removes outlier, reduces redundancy, computes local features, and 
generate vectorization result of the linear structures. Moreover, to provide a 
standard vectorization result for evaluation purposes, we designed and 
implemented a manual tool allowing the user selection of the points of interest 
and generate vectorization results with proper visual feedback. A real-world 
problem of digitizing the location and shape information of the high-voltage 
powerlines is our main task of the pipeline in this thesis and it provides a 
context for analyzing the correctness and effectiveness of each stage in our 
pipeline.


Date:			Friday, 5 June 2020

Time:			2:30pm - 4:30pm

Zoom Meeting:		https://hkust.zoom.us/j/6158493103

Committee Members:	Prof. Long Quan (Supervisor)
 			Prof. Chiew-Lan Tai (Chairperson)
 			Dr. Pedro Sander


**** ALL are Welcome ****