Three-Dimensional Recognition of Solid Objects from a Two-Dimensional Image
Abstract
This thesis addresses the problem of recognizing solid objects in the three-dimensional world, using two-dimensional shape information extracted from a single image. Objects can be partly occluded and can occur in cluttered scenes. A model based approach is taken, where stored models are matched to an image. The matching problem is separated into two stages, which employ different representations of objects. The first stage uses the smallest possible number of local features to find transformations from a model to an image. This minimizes the amount of search required in recognition. The second stage uses the entire edge contour of an object to verify each transformation. This reduces the chance of finding false matches. A new method is developed for computing transformations from a model to an image. It is shown that when perspective viewing is approximated by orthographic projection plus scale, three corresponding model and image points define a unique transformation, up to a reflection. The solution method based on this result only involves second order equations, and thus is fast and robust. Recognizing objects under projection requires features that are relatively stable over changes in viewpoint. Stable features are obtained by segmenting edge contours at zeroes of curvature, because these points are preserved under projection. Each feature defines either a point and an orientation or three points, so only one or two features are needed to compute a transformation. Thus the number of transformations considered in recognition is only quadratic in the number of corresponding model and image features.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1988
- Accession Number
- ADA205667
Entities
People
- Daniel Huttenlocher
Organizations
- Massachusetts Institute of Technology