Abstract
A small bank of correlation filters is designed using an iterative process. The filter bank will recognize continuous views of a selected 3-D target. Filter design begins by finding orthogonal eigenimages that represent the 3-D information about the target. These eigenimages are combined to form the complex valued correlation filters. The set of coefficients used to combine the eigenimages for each filter is determined so that two constraints are simultaneously satisfied. First, each filter must contain all the eigenimages. Second, as the filters are applied to the input scene in sequence, the inner products between the filters and any potential target view must have a constant magnitude. The detection method relies on finding points of constant intensity in the correlation domain as a small bank of filters is applied in sequence.
© 1988 Optical Society of America
PDF ArticleMore Like This
George F. Schils, Ellen Ochoa, and Donald W. Sweeney
MA6 OSA Annual Meeting (FIO) 1987
Kenneth G. Leib and Michael A. Soel
FDD3 OSA Annual Meeting (FIO) 1988
D. W. Sweeney, G. F. Schils, and N. C. Gallagher
ThC2 Signal Recovery and Synthesis (SRS) 1989