A Novel Optoelectronic System for Distortion-Invariant ATR
Abstract
The proposed idea is to enhance existing optical correlation based automatic target recognition (ATR) by introducing scale, translation, and rotation invariance (STRI). Previously, STRI was achieved by utilizing a library of matched filters corresponding to various positions of rotation and scale variation. We find that simple preprocessing can eliminate the need for the library since any STR changes in the input image can be nullified by bringing an input image to what we call its archetype. Physical Optics Corporation's (POC's) new approach includes a segmentation preprocessor (required to select any potential object of interest in an image of a realistic scene), POC's invariant preprocessor, and a standard optical correlator. This combination will greatly increase recognition speed and will simplify the recognition training procedure since only one in-plane position of the target will be needed to provide STRI recognition. The Phase I research described in this report opened a new approach to invariant ATR in which in-phase object position will no longer be of concern and, therefore, the best matched filter can be exploited to attain maximum confidence of identification. This should result in the development of hardware capable of operating in a real-time mode.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1994
- Accession Number
- ADA277985
Entities
People
- Lev Sadovnik