A dissertation submitted for the degree of Master of Science School of Computer Science
University of the Witwatersrand. Johannesburg, 2015. / Rock art is the archaeological term used to describe pre-historic artworks placed on
natural stone and as one of the earliest known traces of modern human creativity, it
is a major component of world history and human heritage. Archival records and the
art itself, however, are rapidly decaying requiring the need to preserve them for future
generations and humanity as a whole. In line with this, the Rock Art Research Institute
digitised their collections of photographs and historical records of the rock art in southern
Africa. This has resulted in the South African Rock Art Digital Archive, a collection of
over 275 000 images of paintings depicting a wide variety of objects such as humans
and animals. The problem with this archive, however, is that most of the images are
not labelled with the objects that appear in them. Manual labelling is infeasible due
to the size of the archive but rock art researchers require this information to perform
text-based search queries. Therefore, in this research, we present the combination of the
Viola Jones object detection framework and a Support Vector Machine to automatically
classify rock art objects. To test it, we have created and assessed the performance of
classi ers for eland, elephant, human, and rhebuck rock art objects. We have performed
the experiments using ve-fold cross-validation and found the results to be promising
considering the variation and deterioration of the paintings.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:wits/oai:wiredspace.wits.ac.za:10539/18583 |
| Date | January 2015 |
| Creators | Purshotam, Amrit |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf, application/pdf |
Page generated in 0.0026 seconds