Kaveh Heidary

Application of supergeneralized matched filters to target classification.

The matched filter (MF) is the optimum linear operator for distinguishing between a fixed signal and noise, given the noise statistics. A generalized matched filter (GMF) is a linear filter that can handle the more difficult problem of a multiple-example signal set, and it reduces to a MF when the signal set has only one member. A supergeneralized matched filter (SGMF) is a set of GMFs and a procedure to combine their results nonlinearly to handle the multisignal problem even better. Obviously the SGMF contains the GMF as a special case. An algorithm for training SGMFs is presented, and it is shown that the algorithm performs quite well even for extremely difficult classification problems.

Superlattice Multinanolayered Thin Films of SiO2/SiO2 + Ge for Thermoelectric Device Applications

Thermoelectric generators convert heat to electricity. Effective thermoelectric materials and devices have a low thermal conductivity and a high electrical conductivity. The performance of thermoelectric materials and devices is shown by a dimensionless figure of merit, ZT = S2σT/K, where S is the Seebeck coefficient, σ is the electrical conductivity, T is the absolute temperature, and K is the thermal conductivity. We have prepared 100 alternating layers of SiO2/SiO2+ Ge superlattice thin films using ion beam–assisted deposition for the thermoelectric generator device application. The 5 MeV Si ion bombardments were performed using the Center for Irradiation Materials’ Pelletron ion beam accelerator to form quantum dots and/or quantum clusters in the multinanolayer superlattice thin films to decrease the cross-plane thermal conductivity and increase the cross-plane Seebeck coefficient and cross-plane electrical conductivity. The thermoelectric and transport properties have been characterized for SiO2/SiO2+ Ge superlattice thin films.

Discrimination among similar looking noisy color patches using Margin Setting

The goal of discrimination of one color from many other similar-appearing colors even when the colored objects show substantial variation or noise is of obvious import. We show how to accomplish that using a technique called Margin Setting. It is possible not only to have very low error rates but also to have some control over the types of errors that do occur. Robust spectral filtering prior to spatial pattern recognition allows subsequent filtering processes to be based on conventional coherent optical correlation that can be done monochromatically.

Thermoelectric properties of SiO2/SiO2+CoSb multi-nanolayered thin films modified by MeV Si ions

We have fabricated the thermoelectric generator devices from 100 alternating layers of SiO2/SiO2+CoSb superlattice thin films using the ion beam–assisted deposition. Rutherford backscattering spectrometry was used for quantitative elemental analysis of Si, Co, and Sb in the multilayer films. The thin films were then modified by 5-MeV Si ion bombardments using the Alabama A&M University Pelletron ion beam accelerator. Quantum dots and/or clusters were produced in the nanolayered superlattice films to decrease the cross-plane thermal conductivity, increase the cross-plane Seebeck coefficient, and the cross-plane electrical conductivity. We have characterized the thermoelectric generator devices before and after Si ion bombardments using the thermoelectric, optical, and surface characterization techniques. The optical absorption amplitude decreased when the first fluence of 1 × 1012 ions/cm2 was introduced from the value of 2.8 to about 1.9 at 200 nm. The figure of merit reached the maximum value of about 0.005 at the fluence of 1 × 1013 ions/cm2.

A unified approach for database analysis and application to ATR performance metrics

An essential component within most approaches used to evaluate ATR algorithm performance is an image database from which are chosen a training set of images. Several fundamental questions arise regarding the adequacy of the database to represent the desired domain of effectiveness, the sufficiency of the training set, potentiality of enhancing the constituents of the training set, suitability to determine signal-to-clutter performance, and realism of fairly comparing performance of ATR algorithms to one another. These questions have been addressed through an investigation into a unified approach for database analysis and how it can be applied to evaluating ATR performance metrics.

Needles in a haystack: Fast spatial search for targets in similar-looking backgrounds

Abstract This paper develops an efficient and robust algorithm that simultaneously detects and locates image anomalies and intrusions. Anomalies refer to image regions that do not belong to expected classes. In situations where most of the image is of one or more types of known background classes while a few isolated regions may belong to unknown classes, the algorithm detects and locates potential intrusions by blanking regions it classifies as members of the known classes. We used a combination of Fourier filtering, a fast linear way to scan the content of the whole scene in parallel, with Margin-Setting, a powerful nonlinear discriminant trained to distinguish members of known classes from everything else. That combination retains the power of Margin-Setting and the simplicity, speed, and locating ability of Fourier filtering. Examples show the ability of this method to remove essentially all background material while leaving the similar looking intrusions intact. The classifier is trained using a few small square patches extracted from images or image regions representing the background classes of interest. Processed images related to four different problems as well as cumulative numerical results of many tests performed on one of those problems are presented. Excellent performance is observed for the examples considered here.

Presmoothing effects in Artificial Color image segmentation

Artificial Color is the application of Natures basic way of discriminating and segmenting images according to their spectra to technologically acquired images. It has proved very successful, but it produces somewhat ragged segmented images when the spectra of the target class and background are so close that some pixels are hard to distinguish even with the most powerful nonlinear discriminants. Post processing, including median filtering or mathematical morphology can improve the resultant segmented images, but they seldom solve the problem entirely. In this paper, we explore application of another technique utilized by Nature - presmoothing. The results are dramatically superior to those obtained by post smoothing. Qualitative and quantitative support for that assertion is offered.

Distortion tolerant correlation filter design

This paper introduces a computationally efficient algorithm for synthesis of a distortion tolerant correlation filter and associated threshold, denoted collectively as the enhanced matched filter (EMF). Application areas of EMF include imagery based automatic target detection and recognition and biometrics. The EMF is synthesized from a set of training images characterizing the target of interest within the expected distortion range. A distinguishing feature of EMF is the ascribed threshold, which is a byproduct of the filter computation process and does not rely on nontarget trainers. The EMF performance is compared to that of the synthetic discriminant function using realistic test scenarios.

Margin-setting with hyperellipsoidal surfaces

In a number of prior papers we described and explored a new pattern recognition method called Margin-Setting that accomplishes excellent generalization using very few training samples. The result was a multi-round classifier with each round consisting of a set of hyperspheres such that if a datum fell within a certain hypersphere, it was labeled with one particular class. Margin-Setting achieves concurrent low Vapnik-Chervonenkis (VC) dimension and high accuracy which is a consequence of partitioning the training set into smaller sets that make this possible. This paper extends Margin-Setting from hyperspheres to hyperellipsoids resulting in improved performance.

Color classification using margin-setting with ellipsoids

This paper presents a robust and computationally efficient genetic algorithm for color classification. It designs well-fitted color space prolate spheroids (ellipsoids) that envelop the training pixels. The ellipsoids are then used to classify unlabeled image pixels in accordance with their color, in order to partition the image. The color classification algorithm described here has very low error rates, boasts very high operational speed, and permits trading higher indecision rates for lower rates of misclassification. The performance of the color classifier developed in this paper is compared with those of the support vector machine (SVM) and the nearest-neighbor (kNN) classifiers. It has been shown that our color classifier outperforms SVM and kNN for partitioning of color images that contain several closely spaced color classes. It has higher correct classification, lower misclassification, and significantly reduced operational latency in comparison with color classifiers based on kNN and SVM.