Surface reconstruction of a blast plate using stereo vision

This thesis presents method for reconstructing and measuring the profile of a blast metal plate. Among the many methods in computer vision, stereo vision using two cameras is chosen as the range finding method in this thesis. This is because it is a non-contact method and hence eliminates the need to calibrate moving parts. A stereo-rig consists of two calibrated cameras and hence gives two view geometry. Stereoscopic reconstruction relies on epipolar geometry to constrain the relationship between the views. The 3-D point is then estimated using triangulation of the corresponding points from the two views. The blast plates that are reconstructed have highly reflective surfaces. This causes a problem due to specular reflection. This thesis further studies the reflective properties of the metal plate surface. Different methods of scanning the plate using the stereo-rig are investigated. The reconstructions obtained from these methods are analyzed for accuracy and consistency. Since low cost cameras are used in constructing the stereo-rig, the point cloud data obtained is further investigated for consistency by aligning different instances of the reconstruction. This is done using the Iterative Closest Programme (ICP) algorithm which tries to align two sets of data iteratively.

Grantor dc:publisher.institution

Department of Electrical Engineering

Year dc:date.issued

2007

Author dc:creator

• Kigozi, Gladys Nabagala

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/14798

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/14798