Target Recognition Using Three Dimensional Laser Range Imagery.
Abstract
This thesis involved the analysis of computer generated synthetic range imagery for the purpose of autonomous target recognition. The scenario is an air to surface missile sensor using a laser rangefinder to image prospective targets and attempt recognition. Synthetic range images of a sophisticated Soviet T 72 tank model were created. Cross correlation was used as a recognition technique. A reference tank image was tested against rotated images and an array of decoys. The reference image was analysed for its most prominent features for the purpose of examining feature extraction as a recognition technique. Two methods of image enhancement were compared: gradient (frequency emphasis) and phase only filtering. It was shown that these two methods exhibited equal performance for recognition of rotated targets, but differently in decoy rejection. Phase only filtering was more effective in the process of discriminating simple decoys from actual tanks. Feature analysis of the model tank revealed its correlation was highly dependent upon what method of image enhancement was used. Keywords: Cross correlations; Feature extraction; Image processing; Phase filtering; Range images.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1986
- Accession Number
- ADA177755
Entities
People
- Richard P. De Fatta
Organizations
- Air Force Institute of Technology