Abstract
A new method for shift- and scale-invariant pattern recognition is based on a modification of the classical matched filtering technique. The nonsymmetrical Fourier spectrum of an input scene matches a filter made from the symmetrical FT of a target. The scale of the nonsymmetrical spectrum of a scene changes with the angle in the Fourier plane. Therefore a prototype filter can match the spectrum of an object having a different size in some direction. To reduce the cross-correlation noise, wedge masks are placed in the Fourier plane of a correlator. A scale search is possible by rotating the wedge masks. In the first variant of the system, a common VanderLugt filter is used and a simple threshold detection of the correlation signals is made in the output plane. In the second, a few circular harmonic filters of a target are used successively and a discrimination decision is made in feature space. Computer experiments were carried out for both variants.
© 1985 Optical Society of America
PDF ArticleMore Like This
Liang Minhua, Wu Shudong, Liu Liren, and Wang Zhijiang
TuI26 Optical Computing (IP) 1989
Henry Stark and Ron Wu
ThD3 Machine Vision (MV) 1985
Y. Sheng and H. H. Arsenault
THE3 OSA Annual Meeting (FIO) 1985