Pattern recognition using binary masks based on the fractional Fourier transform

摘要

This paper presents two new methodologies for pattern recognition invariant to position and rotation. The two methods use the correlation of signatures obtained using a binary rings mask created from the fractional Fourier transform (FRFT). An analysis of fragmented information was performed with 21 images of fossil diatom species. The methodologies were applied to 30 species of phytoplankton to test the performance when rotation distortion is present in a target. Also, a noise analysis was performed. Higher signature correlation values are achieved by working outside the Fourier plane by changing the order of the FRFT.