University of Twente Student Theses

Login

Object-Oriented Analysis of Very High Resolution Orthophotos for Estimating the Population of Slum Areas, Case of Dar-Es-Salaam, Tanzania

Aminipouri, Mehdi (2009) Object-Oriented Analysis of Very High Resolution Orthophotos for Estimating the Population of Slum Areas, Case of Dar-Es-Salaam, Tanzania.

[img] PDF
1MB
Abstract:Unplanned development of urban areas and the creation of slum settlements are responses to rapid population growth and its density in fast growing cities like Dar- Es-Salaam, the former capital of Tanzania. Regarding the scope of the Millennium Development Goals (MDGs) established by United Nations, especially Goal 7 target 11 “the improvement of the quality of at least 100 million slum dwellers by the year 2020”, one of the substantial steps in meeting this goal is to find reliable procedures for detecting and monitoring the slum areas. Hence, obtaining up to date spatial information about informal settlements is of great importance for any decisions that have to be made by urban planners. Use of Very High Resolution (VHR) airborne and satellite imagery in meter or submeter level has generated a new era in proceeding of the information extraction of slum areas based on object-oriented techniques and estimation methods of the number of slum inhabitants. The main objective of this research, therefore, is to determine the feasibility of using VHR orthophotos (0.6 meter pixel size) in creating an accurate inventory of buildings to enable the estimation of slum population based on the extracted building roofs. eCognition software is used for the image segmentation and classification of the objects of interest (roofs in this study). The accuracy of the roof extraction approach and consequently the estimation of slum population are examined on three different study wards called “Charambe”, “Manzese” and “Tandale” based on a sample size of n=550 reference polygons for each site. Two methods of accuracy assessment are utilised: 1. Considering a roof coverage threshold (25%) to take into account the coregistration errors (positional accuracy); and 2. Ignoring the coverage threshold and calculating the classification accuracies and Kappa statistic from the confusion matrices. Applying the first method of accuracy assessment resulted in the followings: In total, 1504 buildings are extracted out of 1650 reference buildings in the study areas which correspond to 91.1% accuracy in the extraction rate. Across all three study sample sites, the total roof area coverage extracted from eCognition software is estimated at 136720 m2 and from reference data 183195 m2. This amount shows that 25.7% of the reference polygons are not covered by the extracted roofs. Thereby, 74.3% of the total population of these three wards is estimated by the applied model.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/92709
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page