A compact photorefractive joint transform correlator for industrial recognition tasks

Abstract

Technology advances in solid state lasers, spatial light modulators and nonlinear optical materials are centrally important for the construction of optoelectronics processors that combine the massive interconnectivity and parallelism of optics with the accuracy and flexibility of digital electronics. In pattern recognition applications, hybrid optical-digital approaches in which optics performs correlation operations and electronics processes the output correlation plane for classification have already been demonstrated(1−2). Today, the performances of semiconductor lasers, diode-pumped YAG lasers, two dimensional liquid crystal light modulators and photorefractive materials allow the introduction of compact and more flexible optical hardware in optoelectronic processors. In this paper, we present a compact and reconfigurable multichannel joint transform optical correlator designed and constructed for industrial recognition applications. The principle of operation is shown in Fig.1. The object to be identified S(x,y) is display on one half of the input scene. The other half of the input, allocated to the reference R(x,y) is split in N subarrays, or channels, each containing a reference object or a calculated version of reference object. The sum R(x,y) + S(x,y) is Fourier transformed and the spectrum is recorded in a dynamic holographic medium. The complex light field produced by reading out the joint-transform power spectrum contains the cross-correlation component R(x,y) ⊗ S (x-2a, y), where 2a is the separation between signal and reference and ⊗ denotes the correlation operation(3). The identification is performed by detecting the position and relative intensities of the correlation peaks in the corresponding subarrays of the output plane.