Mapping illegal dumps in Soweto using object-based Image analysis

Simple item page
dc.contributor.authorSelahle, Sibongiseni
dc.date.accessioned2021-07-02T09:37:27Z
dc.date.available2021-07-02T09:37:27Z
dc.date.issued2020
dc.descriptionA research report submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in partial fulfilment of the requirement for the degree of Master of Science GIS and Remote Sensing at the School of Geography, Archaeology & Environmental Studies, 2020en_ZA
dc.description.abstractThe illegal dumping of waste is one of the major causes of land degradation which presents a problem for the global environment for both the developed and developing world. The illegal dumping of both building rubble and domestic waste results in groundwater pollution, air pollution and the contribution to the greenhouse gas budget. Illegal dumps cost the local municipalities a lot of money for the clean-up and removal of the illegally dumped waste. Therefore, there is a need for mapping and monitoring illegal dumps. Mapping and monitoring illegal dumps will enhance the prevention of the problem and develop technologies to aid the efficiency of service delivery. The traditional methods of detecting and mapping illegal dumps are usually expensive and require a lot of human resources. This research report aimed at developing a remote-sensing based approach to detect and map illegal dumping in Soweto. The objectives of the report were as follows: (I) map and differentiate between illegal domestic waste and the building rubble in Soweto using Worldview-4 imagery and object-based image classification; (II) compare random forest and support vector machine algorithms for classifying the illegal dumping wastes; (III) to identify and analyse the socio-economic factors associated with the spatial distribution of illegal dumping. The results demonstrated the ability of WorldView-4 imagery and object-based image classification techniques to detect, map, and differentiate between building rubble and domestic waste. Machine learning algorithms produced an overall accuracy of 93.98% for Random forest and 94.91% for Support Vector Machine. Additionally, factors such as population density, household size, level of education, and household income were associated with the spatial distribution of these dumps. The high accuracies of the models provide an opportunity for remote sensing tools and techniques such as the ones implemented in this study to be tested for reliability in the greater City of Johannesburg region and other parts of the world as wellen_ZA
dc.description.librarianCK2021en_ZA
dc.facultyFaculty of Scienceen_ZA
dc.identifier.urihttps://hdl.handle.net/10539/31410
dc.language.isoenen_ZA
dc.schoolSchool of Geography, Archaeology & Environmental Studiesen_ZA
dc.titleMapping illegal dumps in Soweto using object-based Image analysisen_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
Sibongiseni Selahle Corrections.pdf
Size:
3.02 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
ETD Collection