Pattern discrimination enhancement using adaptive joint transform correlator.

摘要

This research investigated techniques for improving the performance of joint transform correlators (JTC). Although primarily based on pattern recognition, these techniques may also be applied to general optical signal processing on JTCs.; There were two main approaches to this investigation: (1) to improve the correlation profiles through filtering techniques, and (2) to remove unwanted contents from the joint power spectrum. The former is similar to the inverse filtering technique used in a VanderLugt processor, also branching into three areas: intensity compensation filtering (ICF), spatial synthesis of ICF, and iterative ICF. The latter focused on eliminating the intra-class associations (i.e., all associations between the reference members themselves plus associations among target members). Since these contents offer no benefits but adverse effects, their removal can significantly improve the JTC performance. This was done by subtracting the separate power spectra of the reference and target from the joint power spectrum. As a result, the space bandwidth product and the diffraction efficiency are at least doubled.; The ICF is derived from the inverse power spectrum of the reference such that, when the target matches the reference, the amplitude distribution of the joint power spectrum will be evened by the ICF. Consequently, an impulse will appear around the center of the correlation which indicates a match. The poles in the inverse power spectrum, together with the influence of noise at various strengths, were also investigated. The simulated result shows that the discrimination percent is raised to 70.3% from 36.2% using conventional JTC. Also, when the signal-to-noise level ranging between 0 dB to 40 dB, the correlation peak sharpness is 30 to 60 times better and the diffraction efficiency is at least tripled.; The spatial synthesis technique combined the spatial impulse response of the ICF and the reference function into a spatially synthesized function, which produces a sharp correlation peak output when the target matched the reference.; The iterative technique explored the effects of feedback control on the illumination and composite ICF. The tests were based on recognition of a few similar patterns. The JTC's selectivity was increased through this process. The simulated result shows that a number of resembling characters can not be differentiated or recognized correctly using the conventional JTC but can be done with this feedback technology.; This study is also intended to explore more applications under the current condition. Some future research ideas are suggested in the end of this thesis.