A Neural Network Object Recognition System
Abstract
A system involving neural networks has been developed to identify objects as an instance of a prespecified set of object classes. This system is useful for exploring different neural network configurations. There are three main computation phases of a model based object recognition system: segmentation, feature extraction, and object classification. This report focuses on the object classification stage. For segmentation, a neural network based segmenter developed by M. Tenorio can be used. Other, more conventional segmentation schemes are also available. Neural networks for feature extraction are at an early stage of development; other more conventional techniques have proved to be very fast and effective. Several conventional techniques are available with the current system. Neural network based feature extraction may be added at a later date. The classification stage consists of a very flexible neural network simulator which may be configured for wide range different network concepts. The system is based on VISIX computer vision software environment. It provides an interactive experimentation facility that may be used on a large number of UNIX based workstations. The system has been tested with a number of simple object recognition tasks. (Author) (KR)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1990
- Accession Number
- ADA225627
Entities
People
- A. P. Reeves
- F. P. Kuhl
- R. J. Prokop
Organizations
- United States Army Armament Research, Development and Engineering Center