Carlos J. García Orellana

Acetic acid detection threshold in synthetic wine samples of a portable electronic nose.

Wine quality is related to its intrinsic visual, taste, or aroma characteristics and is reflected in the price paid for that wine. One of the most important wine faults is the excessive concentration of acetic acid which can cause a wine to take on vinegar aromas and reduce its varietal character. Thereby it is very important for the wine industry to have methods, like electronic noses, for real-time monitoring the excessive concentration of acetic acid in wines. However, aroma characterization of alcoholic beverages with sensor array electronic noses is a difficult challenge due to the masking effect of ethanol. In this work, in order to detect the presence of acetic acid in synthetic wine samples (aqueous ethanol solution at 10% v/v) we use a detection unit which consists of a commercial electronic nose and a HSS32 auto sampler, in combination with a neural network classifier (MLP). To find the characteristic vector representative of the sample that we want to classify, first we select the sensors, and the section of the sensors response curves, where the probability of detecting the presence of acetic acid will be higher, and then we apply Principal Component Analysis (PCA) such that each sensor response curve is represented by the coefficients of its first principal components. Results show that the PEN3 electronic nose is able to detect and discriminate wine samples doped with acetic acid in concentrations equal or greater than 2 g/L.

SEGMENTATION OF BOVINE LIVESTOCK IMAGES USING GA AND ASM IN A TWO-STEP APPROACH

In this work a system based on genetic algorithms (GA) and active shape models (ASM) is presented that, in a two-step approach, is able to accurately find a particular object within the image. The aim of the first stage is to approximately locate the object in order to serve as an initialization for the second. Following a method similar to that used by other authors, a model of the desired shape (cows in lateral position) is constructed using PDM, and later the search within the image is carried out based on instances of that model, and using genetic algorithm techniques, for which several objective functions are suggested. In the second stage we attempt to make a fine adjustment of our instance to the object in the image, starting from the result of the previous process. Consequently, we propose two different techniques compared afterwards: deformable contours (specifically active shape models) and a restricted genetic search. Finally, the system is tested over a database of 309 animal images, and results are presented.

NeuSim: A Modular Neural Networks Simulator for Beowulf Clusters

We present a neural net works simulator that we have called NeuSim, designed to work in Beowulf clusters. We have been using this simulator during the last years over several architectures and soon we want to distribute it under GPL license. In this work we offer a detailed description of the simulator, as well as the performance results obtained with a Beowulf cluster of 24 nodes.

GA Techniques Applied to Contour Search in Images of Bovine Livestock

In this work a system based on genetic algorithms is presented that generates valid initializations for deformable models methods. Following a systematics similar to that used by other authors, a model of the shape we are looking for (cows in lateral position) is constructed using PDM, and later the search within the image is made based on instances of that model, and using genetic algorithms techniques. Since we have color images, several objective functions are suggested that take advantage of this information, which are tested later over a database of 309 animal images taken directly in the field.

Adaptive BAM for pattern classification

A new method for the synthesis of neural networks with BAM (bidirectional associative memory) features, based on the ART structure, is presented. Intended for pattern classification, it contains a new procedure for the correct usage of the relation matrix, and avoids the inherent defects of the BAM and its misclassifications with appropriate actions on the thresholds of the neurons of the ART layers. The results clearly indicate that this method leads to a good improvement in the performance that is achievable in a BAM, with a 0% error rate found in a test on the well-known NIST 19 character database.

Application of ANN techniques to automated identification of bovine livestock

In this work a classification system is presented that, taking lateral images of cattle as inputs, is able to identify the animals and classify them by breed into previously learnt classes. The system consists of two fundamental parts. In the first one, a deformable-model-based preprocessing of the image is made, in which the contour of the animal in the photograph is sought, extracted, and normalized. Next, a neural classifier is presented that, supplemented with a decision-maker at its output, makes the distribution into classes. In the last part, the results obtained in a real application of this methodology are presented.

Application of Repeated GA to Deformable Template Matching in Cattle Images

In this work, a system that enables accurate location of contours within outdoors–taken digital images is presented. This system is based on the optimization of an objective function involving a selective edge detector and a distance map extractor, applied sequentially to the original image. In order to guarantee a certain probability of success, the repeated execution of GA is applied. The performance of the system is tested against a database of 138 images of cows in lateral position, obtaining a rate of successes up to 92 % in the best of the cases.

Associative memories for handwritten pattern recognition

We describe the construction of a complete system of handwritten classification on the basis of bidirectional associative memories (BAM). The BAM is synthesized with a two-layer neural network in which the neuron thresholds are adjusted at the moment of learning, and the ideal prototypes for each class are selected using an automatic method. The system is completed with a topological preprocessor that eliminates distortion and noise components. It was tested on the popular NIST#19 database, attaining 100% success rates for all the characters, even under conditions of high levels of contamination due to noise or distortion of the input image.

Application of the new Monte Carlo code AlfaMC to the calibration of alpha-particle sources

Abstract Measurements of α-particle sources require corrections to the counting rate due to scattering and self-absorption in the source and the backing material. In this study, we describe a simple procedure to estimate these corrections using the new Monte Carlo code AlfaMC to consider the effects of scattering and self-absorption conjointly, and so to determine the activity of α emitters. The procedure proposed was applied to 235UO2 sources deposited on highly polished platinum backings. In general, the dependence of the efficiency with source thickness was in good agreement with a simple model considering a linear and a hyperbolic behavior for thin and thick sources, respectively, although significant deviations from this model were found for very thin sources. For these very thin sources, the Monte Carlo simulation revealed to be as a required method in the primary calibration of α-particle sources. The efficiency results obtained by simulation with AlfaMC were in agreement with available efficiency data.

Large neural net simulation under Beowulf-like systems

In this work we broach the problem of large neural network simulation using low-cost distributed systems. We have developed for the purpose high-performance client-server simulation software, where the server runs on a multiprocessor Beowulf system. On the basis of a performance analysis, we propose an estimator for the simulation time.