Elisabet Perez

Dual nonlinear correlator based on computer controlled joint transform processor : digital analysis and optical results

Abstract The hybrid optoelectronic processor, presented in this paper, realizes the dual nonlinear correlation (DNC) in a set-up based on a two-step nonlinear joint transform correlator architecture. In the first step three power spectrum distributions (input scene power spectrum, reference target power spectrum, and the joint power spectrum) necessary for the nonlinear processing are captured with a CCD camera. Nonlinear modification of the joint power spectrum, which does not have to be symmetrical in the input and reference channels, is introduced digitally. In the second step, the modified joint power spectrum is Fourier transformed optically. Numerical analysis of this processor shows a crucial influence of the dynamic range and the limited number of grey levels of the CCD camera during image acquisition in the first step, on the output signal parameters and the discrimination capability of the set-up. Optical results of recognition obtained for noise-free segmented input scenes show that the set-up ...

Optical pattern recognition with adjustable sensitivity to shape and texture

In this paper, an optical pattern recognition system with adjustable sensitivity to shape distortions and texture changes of the objects is presented. Application to a recognition task where the information of texture is the most decisive feature for a given object to be detected is provided. We apply the dual nonlinear correlation (DNC) model along with a support function acting in the frequency domain. This support function performs as an additional nonlinearity that enhances the information of some selected frequency bands related to the textural content of the target. A mathematical analysis allows the authors to show the usefulness of the proposed support function in the frame of the DNC model. The recognition system is applied to accomplish different recognition tasks involving model and real textured objects. The proposed optoelectronic correlator has been used to obtain successful experimental optical results, which are in accordance with the simulated results also provided.

Pattern recognition with variable discrimination capability by dual non-linear optical correlation

The discrimination capability requirements of pattern recognition systems may vary from one given purpose to another. In this work a recognition system with variable and selective discrimination capability is obtained by applying a dual non-linear correlation (DNC) model to a joint-transform correlator. DNC is obtained by means of two non-linear operators that are applied to both the reference and input channels. A particular DNC is given by the values taken by two real control parameters that determine the non-linear operators. In comparison with conventional filtering methods, an increased and variable discrimination capability is achieved by varying the parameters values. Thus, variable tolerances are introduced in the recognition process. Specifically, tolerances to slight shape variations and intensity variations of the objects (alphabetic characters) are analysed in this work. Ranges for the two control parameters are found in each case in order to achieve either an increase or a relaxation in the systems discrimination capability. The developed application is extended to colour pattern recognition by multichannel correlation. In this case, four further applications with selective discrimination capability are developed: pattern recognition with high discrimination capability for shape variations and some tolerance to colour variations and, vice versa, pattern recognition with high discrimination capability for colour variations and some tolerance to slight shape variations; pattern recognition with high discrimination for both shape and colour, and, finally, a tolerance to slight variations in both shape and colour.

Adaptive channel selection for improving chromatic discrimination in colour pattern recognition

Abstract A new method to select colour channels in colour pattern recognition systems is presented to improve the chromatic discrimination capability. Instead of using the conventional RGB decomposition, we propose to use other narrow-band channels that can show more variations of the objects of a scene. This method exploits the spectral information of the objects and is useful in discriminating objects with similar colour. Simulated and experimental optical correlation results, for scenes with objects of similar colours, are presented and discussed. Model and natural objects are considered and a variety of ranges are studied.

Dual nonlinear correlation applied to textured and colour object recognition

Dual nonlinear correlation (DNC) is a general operation in optical pattern recognition involving linear and nonlinear filtering methods. DNC also allows to apply new non-symmetric operators to both the analyzed scene channel and to the reference target channel. A third nonlinearity introduced in the frequency domain allows the control of the region of the spectrum where the DNC is applied. The implementation of the DNC is carried out in a sole filterless optoelectronic processor based on a two-step joint transform correlator assisted by computer. Experimental conditions related to camera and spatial light modulator features have an influence on the method performance. We present some applications of the DNC to textured and color pattern recognition with variable discrimination capability.

Optical angular correlation in fourier domain

Abstract The angular correlation of Fourier spectra is optically implemented by means of a single Fourier transformer. Fourier-domain-based angular correlation, which is a technique specific to periodic pattern recognition and characterization, is efficiently applied in real time to ordinary textile structures. By introducing scale corrections, either isomorphic or anamorphic, the optical system is capable of recognizing different structures of the same sort of fabric even when the fundamental frequencies—or thread densities—do not coincide. Two possible methods to introduce the information into the input image of the optical angular correlator are considered: an opto-mechanical rotator containing a transparency with the input sample image; and an electronic addressed spatial light modulator that displays the input sample image controlled by computer. Experimental results of both possibilities are presented and discussed.

Selective discrimination capability in optical pattern recognition using available SLMs

Optical correlation for pattern recognition with selective and adjustable discrimination capability is based on the Dual Nonlinear Correlation (DNC) algorithm and on a two-step Joint Transform Correlator (JTC) architecture. The DNC encompasses nonlinear power-law processing operating independently on the spectra of the input image and the reference target. But the eventual capabilities of the system strongly depend on some experimental conditions such as quantization, gray-level dynamic range, saturation and other technical characteristics of both the camera and the spatial light modulator used in the JTC. In this work we explore these capabilities for three available modulators (Epson and two different CRLs) acting in both the input and the modified joint power spectra planes for the first and second steps of the JTC, respectively.

Pattern recognition with selective and adjustable discrimination

The discrimination capability requirements of pattern recognition systems may vary depending on the application and the characteristics of the objects to analyse. Our goal is to obtain a single system with a wide and smooth control of its discrimination capability in two senses: the first one, a discrimination capability selective to different aspects of the target (for example, shape, intensity, colour, texture); the second one, a discrimination capability with variable level of sensitivity -or tolerance- to the differences between an object and the target in the aspect selected. This purpose lead us to consider higher levels of complexity in the sensitivity of the recognition system. Thus, we obtain pattern recognition with high discrimination capability for a given aspect along with certain tolerance for another aspect, or vice versa. In this work, we build an optoelectronic system for pattern recognition with selective and adjustable discrimination capability by applying a dual nonlinear correlation model to a joint transform correlator. Dual nonlinear correlation is achieved by means of two nonlinear operators that are applied to both the reference and input channels. A filtering function that limits the region of support in the Fourier plane is additionally introduced. The eventual capabilities of the real system strongly depend on some experimental conditions such as quantization, grey-level dynamic range, saturation and other technical characteristics of both the camera and the spatial light modulator used in the joint transform correlator. We explore these capabilities for a 8-bits and a 12-bits CCD camera and several available modulators. Experimental and simulated results for model and real objects - alphabetic characters, keys and screws- are presented and discussed. They demonstrate that the discrimination capability of the optical recognition system by dual nonlinear correlation can be controlled and adjusted with gradable tolerance to object variations in a given aspect.

Influence of SLM's technical characteristics on correlation results

In this work we evaluate the performance of the dual nonlinear correlator when three available modulators are used. Discrimination capability along with peak-to- correlation energy criteria are used to analyze the experimental correlation planes and recognition results.

Dual nonlinear correlation: optoelectronic realization and application to pattern recognition

Dual nonlinear correlation (DNC) is a general operation in optical pattern recognition involving linear and nonlinear filtering methods. Computer controlled DNC processor is based on a two-step optoelectronic joint transform correlator with the power-law nonlinearities introduced in both channels. The DNC optoelectronic processor is sensitive to the value of power-law nonlinearities and can be adapted to the recognition task on various levels of discrimination capability. A CCD camera and a spatial light modulator are the two basic components of the processor that performs the DNC. Their characteristics such as saturation effect and limited number of quantization levels have strong influence on the correlation signal. In this paper we analyze the performance of the DNC processor and present some applications to textured and color pattern recognition with variable discrimination capability.