Change search
ReferencesLink to record
Permanent link

Direct link
Novel Methodology in Perception of Planar Surfaces in Manhattan World
Blekinge Institute of Technology, School of Engineering.
2008 (English)Independent thesis Advanced level (degree of Master (Two Years))Student thesis
Abstract [en]

Manhattan world are referred to manmade structures with planar surfaces in a scene. In many applications such as robot navigation or mapping is vital to have a 3d perception of such environments. In this thesis a novel methodology is presented to perceive slant surfaces implementing 3D point clouds. Using an enhanced Ncut clustering technique, the point cloud is classified into a number of clusters. Normally in a Manhattan scene one or more of such clusters have a planar nature. To automatically perceive the existents of such plane a series of algorithms is implemented which are consist of ; LS fitting, pruning and RANSAC. Experiments were carried out in MATLAB for both simulated data and real world data. 5 scenes were simulated with 10 noise level in a way that they would resemble the real world data. In addition, Prime Scene sensor was used to collect data from 5 different scenes. In total, the algorithm was tested on 100 point clouds. According to the obtained results, the proposed methodolgy was able to successfully extract planar surfaces in each scene. The performance of the enhanced Ncut algorithm is compared with previous methods of K-means and the original Ncut where each of them separately used in the methodology. The results confirmed significant improvement of the enhanced Ncut over the original Ncut and K-means method. Also our finding has already been published as a conference paper and it has been submitted for a journal publication. These can be seen as good indicators for novelty of the methodology.

Place, publisher, year, edition, pages
2008. , 78 p.
Keyword [en]
Plane Extraction, 3D mapping, Point Cloud, Normalized Cut, Clustering, RANSAC
National Category
Signal Processing
URN: urn:nbn:se:bth-5250Local ID: diva2:832620
Available from: 2015-04-22 Created: 2012-06-13 Last updated: 2015-06-30Bibliographically approved

Open Access in DiVA

fulltext(2622 kB)19 downloads
File information
File name FULLTEXT01.pdfFile size 2622 kBChecksum SHA-512
Type fulltextMimetype application/pdf

By organisation
School of Engineering
Signal Processing

Search outside of DiVA

GoogleGoogle Scholar
Total: 19 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 15 hits
ReferencesLink to record
Permanent link

Direct link