Abstract

A method for recognition of objects with undefined shapes is proposed for use in multispectral analysis. The method involves an orthogonalization procedure in an extended multidimensional spectral range along with separate utilization of information on object brightness. An optimization procedure is performed to reduce the numbers of two types of recognition errors. This method demonstrated a substantial advantage in terms of precision of recognition over the least square method, on which most recognition methods are based.

© 2020 Optical Society of America

Method of synthesized phase objects for pattern recognition: matched filtering

Pavel V. Yezhov, Alexander V. Kuzmenko, Jin-Tae Kim, and Tatiana N. Smirnova
Opt. Express 20(28) 29854-29866 (2012)

Remote object recognition by analysis of surface structure

J. Wurster, H. Stark, E. T. Olsen, and K. Kogler
J. Opt. Soc. Am. A 12(6) 1242-1253 (1995)

Recognition of threaded objects by spatial spectrum analysis

Bryan J. Stossel and Nicholas George
Appl. Opt. 30(11) 1379-1387 (1991)

Three-dimensional object-distortion-tolerant recognition for integral imaging using independent component analysis

Cuong Manh Do, Raúl Martínez-Cuenca, and Bahram Javidi
J. Opt. Soc. Am. A 26(2) 245-251 (2009)

Microwave projection imaging for refractive objects: a new method

K. I. Schultz and D. L. Jaggard
J. Opt. Soc. Am. A 4(9) 1773-1782 (1987)