Optical Data Processing for Missile Guidance.
Abstract
The initial phase of our synthetic discriminant function optical pattern recognition research has been completed. This hyperspace description allows the use of off-line synthesis of a filter function that provides both intra-class recognition (of a key object independent of geometrical distortions) and inter-class discrimination (of a key object from candidate false targets). A new theoretical model for multi-sensor data, a description of a general multi-sensor processor and a statistical analysis of both have also been completed. Initial experimental correlations were obtained and found to be in agreement with our model and theoretical analysis. A new parameter, the correlation measure, was found to be of use in determining the multi-sensor nature of input imagery and a (GRAD) digital preprocessing operator was found to be quite adequate for the cases considered. Our initial optical correlations on the autonomous terminal homing data base were also completed. In this research, aperture correlation effects were overcome by a new technique, optical and digital correlations on the same data base were obtained and compared (with a high pass filter model for the optical matched filter found to be adequate and appropriate for many applications), and sensed/synthetic optical image correlations were obtained (with the help of photoreduced imagery and advanced preprocessing operations). (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1981
- Accession Number
- ADA107592
Entities
People
- David P. Casasent
Organizations
- Carnegie Mellon University