Integrated optical processors in pattern recognition

Abstract

Integrated coherent optical processors are presented and their applications to the field of pattern recognition are considered. It is noted that having coherent optical processors in integrated forms removes many constraints on field applications of conventional optical processors. However, integration introduces some new problems that could degrade optical processor performance. The effects of first-order sources of degradation on the optical architectures proposed here are investigated. How they would affect the processing algorithms developed here is also reported. The problem of preprocessing remotely sensed data to achieve high data rate reduction is addressed. Two distance measure optical processors that operate on the principle of interferometry are proposed. Processing algorithms (parametric and nonparametric) that take full advantage of the statistical dependence of the sensed data on their neighbors as well as the high processing speed of the optical processors are developed. System performance is evaluated and it is shown that intensity fluctuations of the optical source (laser) can have an adverse effect on the entire system.