Towards Change Detection for Laser Scanning Point Clouds

Obtaining an unprecedented amount of topographic data is easier than ever. There has been very extensive research on the analysis of satellite and aerial images, however automatic change-detection and geo-database updates remain challenging tasks. The goal of this thesis is to develop a computational framework to process LiDAR (Light Detection and Ranging) data for disaster response and rapid recovery in near real-time. The specific objectives are the following: (i) develop preprocessing that will enable near real-time processing; (ii) develop an algorithm for water-detection; (iii) develop an algorithm for change-detection; (iv) develop an algorithm capable of classifying the change; and (v) evaluate the performance of these methods on representative datasets. We propose a method for each of the objectives. Specifically, near real-time water- and change-detection and classification on 3D point clouds. Our algorithm addresses the issue of the high dimensionality of the data by formulating the problem as an approximate nearest neighbour problem in a local frame. We have applied the algorithm to the 2002 and 2010 point cloud datasets of Galveston Island and have been able to detect and classify water, buildings, low, and high vegetation change. This analysis of LiDAR data is important as it provides a unique ability to measure high-resolution 3D change and can provide accurate information to emergency response units in natural disasters such as earthquakes, hurricanes, and tsunami.Computer Science, Department o