Automatic Recognition of Complex Three-Dimensional Objects from Optical Images,
Abstract
The paper surveys techniques which have been used for automatic recognition of aircraft and other complex three-dimensional objects from optical images. All methods reported in unclassified literature are described and compared. A detailed description of an experimental system using live television images is included. This system makes use of certain features derived from the central moments of a binary image of the object. These features, called moment invariants are invariant with image rotation, translation, and size. It is reported that a classification error rate of only five percent was obtained in a laboratory experiment involving noisy images of six aircraft types. Human error rates in classifying the same images averaged ten percent. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1973
- Accession Number
- AD0774103
Entities
People
- Robert B. Mcghee
Organizations
- Ohio State University