Rotation- and size-invariant object identification system using the liquid crystal on silicon device

Abstract

We propose a shape discrimination technique using the optical matched spatial filter, which gives an identical result against randomly scattered objects of the same shape but different rotation and scale change. The main idea in this study is to apply optical coordinate transformation, which transforms rotation and scale change including a linear shift into translation, not to the objects themselves but to their Fourier spectra. Thus, rotation and scale change as well as the linear shift can be processed with the same optical matched spatial filter. In addition, we built a two-wavelength system for simultaneous processing of the optical matched spatial filter and optical coordinate transformation. In this system, both Fourier transform and the coordinate transformation were implemented merely by displaying computer-generated holograms (CGH) on the liquid crystal on silicon (LCOS) devices, resulting in a compact lensless system.