A trainable freeform surface-based registration system for rigid three-dimensional objects

Ray Hidayat (PhD Student)

Dept Computer Science and Software Engineering, University of Canterbury

Wed Mar 19 10:00:00 NZDT 2008 in Room 315, Erskine Building

Abstract

Vision is a highly important way of gathering information as it allows complex interaction with the real world. The long-term goal of this project is to construct an algorithm which can recognise any rigid three-dimensional object, given training. To constrain the problem, the object must be opaque and non-reflective. The current plan to achieve this goal is to recognise the surface features that comprise an object, which will be identified by shape and texture analysis. This talk will focus on the techniques that are used to extract shape from an image, including techniques such as Gabor segmentation, Canny edge detection, Quadrature Disambiguation and a new method called Papari boundary detection. From there an overview of how shape analysis fits into a full recognition system will be described.