University of Twente Student Theses

Login

Deep Learning-based Change Detection and Classification for Airborne Laser Scanning Data

Nofulla, Jorges (2023) Deep Learning-based Change Detection and Classification for Airborne Laser Scanning Data.

[img] PDF
7MB
Abstract:In this Master's thesis, we propose a novel approach to urban change detection on raw point clouds, utilizing and evaluating Random Forest (RF), Fully Convolutional Neural Networks (FCNN), and Convolutional Neural Networks (CNN) models. We apply these models to two distinct datasets, a simulated Urb3DCD dataset and a real-world AHN dataset from the Netherlands. Our research builds upon successful 2D change detection methods, adapting them for the 3D domain to provide a simplified, yet efficient and precise, detection method. The performance of our models underscores the importance of feature selection, the quality and representativeness of the training data, and the ability of models to understand spatial relationships. While each model has unique strengths, with the RF model performing particularly well in well-represented areas and deep learning models effectively differentiating between similar classes, the CNN model enhances accuracy by incorporating spatial relationships. We highlight the potential of simpler methods for processing point cloud data and emphasize the challenges when applied in real-world scenarios. This work not only expands the current understanding of urban change detection using 3D point cloud data but also paves the way for future research in this field.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Subject:38 earth sciences, 54 computer science
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/95642
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page