Student:	Hayrettin ACAR
Supervisor:	Prof. Dr. Fevzi KARSLI
Department:	Harita Mühendisliği
Institution:	Graduate School of Natural and Applied Sciences
University:	Karadeniz Technical University Turkey

Title of the Thesis:	AUTOMATIC DETECTION OF BUILDING ROOF POINTS WITH THE 3D POINT CLOUD DATA PRODUCED FROM HIGH RESOLUTIONS IMAGES AND TRUE ORTHOPHOTO GENERATION
Level:	Ph.D.
Acceptance Date:	5/7/2018
Number of Pages:	103
Registration Number:	Di1266

Summary:

In recent years, 3D point cloud data generation has generally been achieved by laser

scanning systems, also the creation and use of the generated 3D point cloud data with the

image matching increased rapidly. The automatic extraction of building boundaries from 3D

point clouds generated by using images is very important for real estate valuation, city

planning and modeling, cadastral applications and updates, defense, natural disaster

management and many other areas. In this thesis, only high resolution aerial photographs

were used for the automatic extraction of building roof points from 3D dense point cloud

data generated for Vaihingen and Ordu study area and the true orthophoto production was

investigated with the improved/precise digital elevation model (DEM) produced later. As a

result, the points of the building roofs are automatically detected by the new algorithm

proposed in the 3D point cloud data and the true orthophotos of these zones are produced.

The Correctness, Completeness and Quality metrics that were calculated in the accuracy

analysis tests were 94.7%, 98.1%, and 93.1% for the Vaihingen test area and 98.1%, 95.1%,

and 94.1% for the Ordu test area, respectively.

Keywords: Building extraction, Image matching, 3D point cloud, Semi global matching