Abstract

In our early study, we have developed an apparatus for optically and automatically recognizing particular shapes in the image. This system has a coordinate transformation processing and matched filtering, and can recognize the shape even if the size and the rotational angle of the target shape change. Furthermore, We achieved lensless optical computing by using the liquid crystal on silicon (LCoS) device, and could be carried out at the same time matched filter and coordinate transformation processing by using different wavelengths of light. Lensless matched filter and optical coordinate transformation processing has been achieved by displaying computer-generated hologram (CGH) on the LCoS. The optical matched filtering has the shift invariant characteristic, which the recognition performance does not change for object translation. However, recognition ability significantly degrades by changes in rotational angle and size. Therefore, rotation and size invariant including shift invariant optical shape recognition system was acheived by the optical coordinate transformation that converts rotation and scale variation into translation. However, this system can be used only to amplitude objects, which have only an amplitude transmittance.

© 2014 Japan Society of Applied Physics, Optical Society of America