Uri Mahlab

Genetic algorithm for optical pattern recognition.

A genetic algorithm is used to generate binary reference functions for optical pattern recognition and classification. Procedures based on the properties of convex functions can be implemented directly on hybrid electro-optical systems. Computer simulations demonstrate the efficiency of this novel approach.

Entropy optimized filter for pattern recognition.

Properties of the entropy function encountered in physics and information theory are employed in the generation of highly selective spatial filters for pattern recognition. Computer simulations and laboratory experiments demonstrate efficient recognition of single patterns or classes even when these are submerged in high level random noise.

Iterative optimization algorithms for filter generation in optical correlators : a comparison

A general approach to the implementation of highly selective spatial filters for pattern recognition leads to a nonlinear optimization problem. Three optimization algorithms, hill climbing (direct search), simulated annealing, and the genetic algorithm, were investigated for implementation on hybrid electro-optical systems. Experimental results indicate the substantial superiority of the genetic algorithm in terms of operating speed and performance quality.

Iterative generation of complex reference functions in a joint-transform correlator.

Iterative learning procedures on hybrid electro-optic systems are employed to generate complex discriminant reference functions. The procedures are implemented experimentally on a joint-transform correlator by using a single inexpensive spatial light modulator. High-quality class discrimination is demonstrated even in the presence of noise.

Target location measurement by optical correlators: a performance criterion

When optical correlators are used for target location measurement or tracking, errors in the position estimation may be induced by the presence of noise and additional objects in the field of view. The variance of the positioning error is evaluated with its dependence on the capabilities of the spatial filter to discriminate among the patterns in the observation field, the shape of the correlation functions, and the noise.

Optical pattern recognition based on convex functions

Spatial filters based on the properties of convex functions can be generated directly on a hybrid electro-optical system. As an example, simulated annealing was used to design highly selective spatial filters. Laboratory experiments demonstrated efficient pattern recognition and class discrimination. Several procedures for designing synthetic discriminant functions discussed in the literature are shown to be special cases of the present procedure.

Iterative generation of holograms on spatial light modulators.

Iterative learning procedures on a hybrid electro-optic system can be employed to generate holograms on inexpensive liquid-crystal-television spatial light modulators. The algorithm takes into account random electronic noise in the system and compensates for spatial-light-modulator distortions. Experimental results are given for the reconstruction of intensity distribution, and computer complex amplitude reconstruction.

Error probability in optical pattern recognition

Abstract The operation of spatial filters in the presence of noise is analyzed in terms of detection error probability and discrimination. While in a theoretical situation the classical matched filter is optimal, if practical limitations ar taken into account, various synthetic filters may have better performance. This is shown to be the case, in particular, for the entropy optimized filter.

Adaptive learning with joint transform correlators

Iterative learning procedures can be employed to generate synthetic reference patterns for joint transform correlators. These patterns can be implemented directly on inexpensive spatial light modulators (SLMs) based on liquid crystal television sets. The algorithm takes into account random electronic noise in the system and compensates for SLM distortions. Experimental results demonstrate the advantages of the new approach with two algorithmic examples.

Implementation of non-linearity management for Gaussian pulses in a fiber-optic link by means of second-harmonic-generating modules

Abstract We introduce a model for the non-linearity compensation (“management”), based on second-harmonic-generating [ χ (2) ] elements, integrated with amplifiers, which are periodically inserted into a fiber-optic link. The link features the Kerr [ χ (3) ] non-linearity, group-velocity dispersion (GVD), and loss. Simulations demonstrate that the system can directly provide for effective compensation of the fiber non-linearity for quasi-Gaussian pulses over the propagation distance z up to 10 fiber spans between the χ (2) modules, and the interaction between adjacent pulses does not manifest itself over the same propagation distance, provided that the separation between them is ≃5 widths of the pulse or more. The results are sensitive to the correct choice of the mismatch parameter in the χ (2) module, but not to variation of other parameters (for instance, peak power of the pulses). For larger values of z , distortion of the pulses commences under the action of GVD. This can be prevented by periodic dispersion compensation. Although the latter is not directly included into the model, it is demonstrated that the non-linearity and dispersion compensation can be implemented independently, as they take place at essentially different scales of z for typical NRZ pulses. Thus, the stable-transmission distance can be expanded much farther.