EuroGP2001 18-20 April 2001
Lake Como (Milan), Italy
home | post conference information | papers | EvoWorkshops | committees   





EvoIASP abstracts

Genetic Snakes for Color Images Segmentation

Lucia Ballerini

The world of meat faces a permanent need for new methods of meat quality evaluation. Recent advances in the area of computer and video processing have created new ways to monitor quality in the food industry. In this paper we propose a segmentation method to separate connective tissue from meat. We propose the use of Genetic Snakes, that are active contour models, also known as snakes, with an energy minimization procedure based on Genetic Algorithms (GA). Genetic Snakes have been proposed to overcome some limits of the classical snakes, as initialization, existence of multiple minima, and the selection of elasticity parameters, and have both successfully applied to medical and radar images. We extend the formulation of Genetic Snakes in two ways, by exploring additional internal and external energy terms and by applying them to color images. We employ a modified version of the image energy which considers the gradient of the three color RGB (red, green and blue) components. Experimental results on synthetic images as well as on meat images are reported. Images used in this work are color camera photographs of beef meat.

A Distributed Genetic Algorithm for Parameters Optimization to Detect Microcalcifications in Digital Mammograms

Alessandro Bevilacqua, Renato Campanini, Nico Lanconelli

In this paper, we investigate the improvement obtained by applying a distributed genetic algorithm to a problem of parameter optimization in medical images analysis. We setup a method for the detection of clustered microcalcifications in digital mammograms, based on statistical techniques and multiresolution analysis by means of wavelet transform. The optimization of this scheme requires multiple runs on a set of 40 images, in order to obtain relevant statistics. We aim to evaluate how fluctuations of some parameters values of the detection method influence the performance of our system. A distributed genetic algorithm supervising this process allowed to improve of some percents previous results obtained after having ``hand tuned'' these parameters for a long time. At last, we have been able to find out parameters not influencing performance at all.

Dynamic Flies: Using Real-Time Parisian Evolution in Robotics

Amine M. Boumaza and Jean Louchet

The Fly algorithm is a Parisian evolution strategy devised for parameter space exploration in computer vision applications, which has been applied to stereovision. The resulting scene model is a set of 3-D points which concentrate upon the surfaces of obstacles. In this paper, we present how the evolutionary scene analysis can be continuously updated and integrated into a specific real-time mobile robot navigation system. Simulation-based experimental results are presented.

ARPIA: a High-Level Evolutionary Test Signal Generator

Fulvio Corno, Gianluca Cumani, Matteo Sonza Reorda and Giovanni Squillero

The integrated circuits design flow is rapidly moving towards higher description levels. However, test-related activities are lacking behind this trend, mainly since effective fault models and test signals generators are still missing. This paper proposes ARPIA, a new simulation-based evolutionary test generator. ARPIA adopts an innovative high-level fault model that enables efficient fault simulation and guarantees good correlation with gate-level results. The approach exploits an evolutionary algorithm to drive the search of effective patterns within the gigantic space of all possible signal sequences. ARPIA operates on register-transfer level VHDL descriptions and generates effective test patterns. Experimental results show that the achieved results are comparable or better than those obtained by high-level similar approaches or even by gate-level ones.

A Pursuit Architecture for Signal Analysis

Adelino R. Ferreira da Silva

One of the main goals of signal analysis has been the development of signal representations in terms of elementary waveforms or atoms. Dictionaries are collections of atoms with common parameterized features. We present a pursuit methodology to optimize redundant atomic representations from several dictionaries. The architecture exploits notions of modularity and coadaptation between atoms, in order to evolve an optimized signal representation. Modularity is modeled by dictionaries. Coadaptation is promoted by introducing self-adaptive, gene expression weights associated with the genetic representation of a signal in a proper dictionary space. The proposed model is tested on atomic pattern recognition problems.

Genetic Algorithm Based Heuristic Measure for Pattern Similarity in Kirlian Photographs

Mario Koppen, Bertram Nickolay and Hendrik Treugut

This paper presents the use of a genetic algorithm based heuristic measure for quantifying perceptable similarity of visual patterns by the example of Kirlian photographs. Measuring similarity of such patterns can be considered a trade-off between quantifying strong similarity for some parts of the pattern, and the neglection of the accidental abscense of other pattern parts as well. For this reason, the use of a dynamic measure instead of a static one is motivated. Due to their well-known schemata processing abilities, genetic algorithm seem to be a good choice for ``performing'' such a measurement. The results obtained from a real set of Kirlian images shows that the ranking of the proposed heuristic measure is able to reflect the apparent visual similarity ranking of Kirlian patterns.

Evolutionary Signal Enhancement Based on Holder Regularity Analysis

Jacques Levy Vehel and Evelyne Lutton

We present an approach for signal en-han-ce-ment based on the analysis of the local H"{o}lder regularity. The method does not make explicit assumptions on the type of noise or on the global smoothness of the original data, but rather supposes that signal en-han-ce-ment is equivalent to increasing the ho regularity at each point. The problem of finding a signal with prescribed regularity that is as near as possible to the original signal does not admit a closed form solution in general. Attempts have been done previously on an analytical basis for simplified cases cite{jlv2}. We address here the general problem with the help of an evolutionary algorithm. Our method is well adapted to the case where the signal to be recovered is itself very irregular, e.g. nowhere differentiable with rapidly varying local regularity. In particular, we show an application to SAR image denoising where this technique yields good results compared to other algorithms. The implementation of the evolutionary algorithm has been done using the EASEA (EAsy specification of Evolutionary Algorithms) language.

Building ARMA Models with Genetic Algorithms

Tommaso Minerva and Irene Poli

The current state of the art in selecting ARMA time series models requires competence and experience on the part of the practitioner, and sometimes the results are not very satisfactory. In this paper, we propose a new automatic approach to the model selection problem, based upon evolutionary computation. We build a genetic algorithm which evolves the representation of a predictive model, choosing both the orders and the predictors of the model. In simulation studies, the procedure succeeded in identifying the data generating process in the great majority of cases studied.

Evolving Market Index Trading Rules using Grammatical Evolution

Michael O'Neill, Anthony Brabazon, Conor Ryan and J.J. Collins

This study examines the potential of an evolutionary automatic programming methodology to uncover a series of useful technical trading rules for the UK FTSE 100 stock index. Index values for the period 26/4/1984 to 4/12/1997 are used to train and test the model. The preliminary findings indicate that the methodology has much potential, outperforming the benchmark strategy adopted.

Autonomous Photogrammetric Network Design using Genetic Algorithms

Gustavo Olague

This work describes the use of genetic algorithms for automating the photogrammetric network design process. When planning a photogrammetric network, the cameras should be placed in order to satisfy a set of interrelated and competing constraints. Furthermore, when the object is three-dimensional a combinatorial problem occurs. Genetic algorithms are stochastic optimization techniques, which have proved useful at solving computationally difficult problems with high combinatorial aspects. EPOCA (an acronym for ``Evolving POsitions of CAmeras'') has been developed using a three-dimensional CAD interface. EPOCA is a genetic based system that provides the attitude of each camera in the network, taking into account the imaging geometry, as well as several major constraints like visibility, convergence angle, and workspace constraint. EPOCA reproduces configurations reported in the photogrammetric literature. Moreover, the system can design networks for several adjoining planes and complex objects opening interesting new research avenues.

The Biological Concept of Neoteny in Evolutionary Colour Image Segmentation: Simple Experiments in Simple Non-Memetic Genetic Algorithms

Vitorino Ramos

Neoteny, also spelled Paedomorphosis, can be defined in biological terms as the retention by an organism of juvenile or even larval traits into later life. In some species, all morphological development is retarded; the organism is juvenilized but sexually mature. Such shifts of reproductive capability would appear to have adaptive significance to organisms that exhibit it. In terms of evolutionary theory, the process of paedomorphosis suggests that larval stages and developmental phases of existing organisms may give rise, under certain circumstances, to wholly new organisms. Although the present work does not pretend to model or simulate the biological details of such a concept in any way, these ideas were incorporated by a rather simple abstract computational strategy, in order to allow (if possible) for faster convergence into simple non-memetic Genetic Algorithms, i.e. without using local improvement procedures (e.g. via Baldwin or Lamarckian learning). As a case-study, the Genetic Algorithm was used for colour image segmentation purposes by using K-mean unsupervised clustering methods, namely for guiding the evolutionary algorithm in his search for finding the optimal or sub-optimal data partition. Average results suggest that the use of neotonic strategies by employing juvenile genotypes into the later generations and the use of linear-dynamic mutation rates instead of constant, can increase fitness values by 58% comparing to classical Genetic Algorithms, independently from the starting population characteristics on the search space.

Using of Evolutionary Computations in Image Processing for Quantitative Atlas of Drosophila Genes Expression

Alexander V. Spirov, Dmitry L. Timakin, John Reinitz and David Kosman

It is well known, that organism of animal, consisting of many billions cells, is formed by consequent divisions of the only cell - zygote. In so doing, embryo cells are permanently communicating by means of biochemical signals. As a result, proper genes were being activated at proper time in proper cells of the embryo.
Modern confocal microscopes being equipped by lasers and computers give possibility to trace-through the cell fate of early embryo for such a classical model object, as fruit fly {it Drosophila melanogaster}. By this approach, it is possible to retrace the detailed dynamics of activity of genes-controllers of development with the resolution on the level of individual nuclei for each of 4-6 thousand cells, composing early fly embryo.
The final result of this analysis will be the quantitative atlas of {it Drosophila} genes action (expression): To achieve this aim we need to receive statistically authentic summary picture of detailed pattern dynamics proceeding from a large number of scanned embryos. This presupposes the elaboration of the methods of preprocessing, elastic deformation, registration and interpolation of the confocal-microscopy images of embryos.
For this purpose we apply modern heuristic methods of optimization to the processing of our images. Namely classic GA approach is used for finding a suitable elastic deformation, for registering the images and for finding a Fourier interpolation of concentration (gene-expression) surfaces. All GA programs considered are the developments of ''evolution strategies program'' from EO-0.8.5 C++ library (Merelo).

Web site designed and maintained by Chris Osborne at EvoNet.