Pseudo Homotopy Trees for Unsupervised Building Detection in Irregularly Spaced LiDAR, FIT Lecture PDF Print E-mail
Written by Nicholas Shorter   
Sunday, 25 October 2009 14:11

On Friday, October 30, from noon to 1pm, Nicholas was invited by Dr. Georgios Anagnostopulos to give a lecture at the Florida Institute of Technology (FIT) about his dissertation research, specifically his work with unsupervised building detection from Irregular LiDAR data

This lecture was recorded and has been hosted as follows:


A PDF Version of the presentation is available via <this link>.

A PDF Version of the flier distributed to advertise the talk is provided via <this link>.

This lecture has been posted to YouTube in 5 parts: (I), (II), (III), (IV), (V)


The abstract for the presentation is as follows:

A novel paradigm is presented which employs homotopy trees, a branch of topography, for building detection in Light Detection and Ranging (LiDAR) data.  The method is developed for irregularly spaced LiDAR data and therefore it can also be applied to rasterized/grid spaced data without any modifications.  Using features extracted from either the first and/or last returns (if available) of the LiDAR pulses and the triangulation of the LiDAR data, the proposed method can discriminate ground from non-ground points and subsequently differentiate non-ground as building or non-building points.  The method is unsupervised, no training phases are necessary.  The only assumption the algorithm makes about the buildings is that they exist as structures which protrude from the ground with a minimal predefined area and height and have a roof.  Results are provided for two different actual data sets without changing any of the algorithm's parameters.


Last Updated on Sunday, 02 October 2011 16:37