Kerstin Malmqvist

Soft combination of neural classifiers: a comparative study

This paper presents four schemes for soft fusion of the outputs of multiple classifiers. In the first three approaches, the weights assigned to the classifiers or groups of them are data dependent. The first approach involves the calculation of fuzzy integrals. The second scheme performs weighted averaging with data-dependent weights. The third approach performs linear combination of the outputs of classifiers via the BADD defuzzification strategy. In the last scheme, the outputs of multiple classifiers are combined using Zimmermanns compensatory operator. An empirical evaluation using widely accessible data sets substantiates the validity of the approaches with data-dependent weights, compared to various existing combination schemes of multiple classifiers.

Colour image segmentation by modular neural network

In this paper segmentation of colour images is treated as a problem of classification of colour pixels. A hierarchical modular neural network for classification of colour pixels is presented. The n ...

Neural networks based colour measuring for process monitoring and control in multicoloured newspaper printing

This paper presents a neural networks based method and a system for colour measurements on printed halftone multicoloured pictures and halftone multi-coloured bars in newspapers. The measured values, called a colour vector, are used by the operator controlling the printing process to make appropriate ink feed adjustments to compensate for colour deviations of the picture being measured from the desired print. By the colour vector concept, we mean the CMY or CMYK ( c yan, m agenta, y ellow, and blac k ) vector, which lives in the three- or four-dimensional space of printing inks. Two factors contribute to values of the vector components, namely the percentage of the area covered by cyan, magenta, yellow and black inks (tonal values) and ink densities. Values of the colour vector components increase if tonal values or ink densities rise, and vice versa. If some reference values of the colour vector components are set from a desired print, then after an appropriate calibration, the colour vector measured on an actual halftone multicoloured area directly shows how much the operator needs to raise or lower the cyan, magenta, yellow and black ink densities to compensate for colour deviation from the desired print. The 18 months experience of the use of the system in the printing shop witnesses its usefulness through the improved quality of multicoloured pictures, the reduced consumption of inks and, therefore, less severe problems of smearing and printing through.

A New Method for Colour Measurements in Graphic Arts

This article presents a method for colour measurements directly on printed half-tone multicoloured pictures. The article introduces the concept of colour impression. By this concept we mean the CMY ...

Monitoring the de-inking process through neural network-based colour image analysis

This paper presents an approach to determining the colours of specks in an image of a pulp being recycled. The task is solved through colour classification by an artificial neural network. The network is trained using fuzzy possibilistic target values. The number of colour classes found in the images is determined through the self-organising process in the two-dimensional self-organising map. The experiments performed have shown that the colour classification results correspond well with human perception of the colours of the specks.

Using Unlabelled Data to Train a Multilayer Perceptron

This Letter presents an approach to using both labelled and unlabelled data to train a multilayer perceptron. The unlabelled data are iteratively pre-processed by a perceptron being trained to obtain the soft class label estimates. It is demonstrated that substantial gains in classification performance may be achieved from the use of the approach when the labelled data do not adequately represent the entire class distributions. The experimental investigations performed have shown that the approach proposed may be successfully used to train neural networks for learning different classification problems.

Colour classification by neural networks in graphic arts

This paper presents a hierarchical modular neural network for colour classification in graphic arts, capable of distinguishing among very similar colour classes. The network performs analysis in a rough to fine fashion, and is able to achieve a high average classification speed and a low classification error. In the rough stage of the analysis, clusters of highly overlapping colour classes are detected. Discrimination between such colour classes is performed in the next stage by using additional colour information from the surroundings of the pixel being classified. Committees of networks make decisions in the next stage. Outputs of members of the committees are adaptively fused through the BADD defuzzification strategy or the discrete Choquet fuzzy integral. The structure of the network is automatically established during the training process. Experimental investigations show the capability of the network to distinguish among very similar colour classes that can occur in multicoloured printed pictures. The classification accuracy obtained is sufficient for the network to be used for inspecting the quality of multicoloured prints.

SELECTING NEURAL NETWORKS FOR A COMMITTEE DECISION

To improve recognition results, decisions of multiple neural networks can be aggregated into a committee decision. In contrast to the ordinary approach of utilizing all neural networks available to make a committee decision, we propose creating adaptive committees, which are specific for each input data point. A prediction network is used to identify classification neural networks to be fused for making a committee decision about a given input data point. The jth output value of the prediction network expresses the expectation level that the jth classification neural network will make a correct decision about the class label of a given input data point. The proposed technique is tested in three aggregation schemes, namely majority vote, averaging, and aggregation by the median rule and compared with the ordinary neural networks fusion approach. The effectiveness of the approach is demonstrated on two artificial and three real data sets.

Detecting and measuring rings in banknote images

Various intelligent systems show a rapidly growing potential use of visual information processing technologies. This paper presents an example of employing visual information processing technologies for detecting and measuring rings in banknote images. The size of the rings is one of parameters used to inspect the banknote printing quality. The approach developed consists of two phases. In the first phase, based on histogram processing and data clustering, image areas containing rings are localized and edges of the rings are detected. Then, in the second phase, applying the hard and possibilistic spherical shell clustering to the extracted edge pixels the ring centre and radii are estimated. The experimental investigations performed have shown that even highly occluded rings are robustly detected. Several prototypes of the system developed have been installed in two banknote printing shops in Europe.

Using unlabeled data for learning classification problems

This chapter presents an approach of using unlabeled data for learning classification problems. The chapter consists of two parts. In the first part of the chapter, an approach of using both labeled and unlabeled data to train a multilayer percetron is presented. The approach banks on the assumption that regions of low pattern density usually separate data classes. The unlabeled data are iteratively preprocessed by a perceptron being trained to obtain the soft class label estimates. It is demonstrated that substantial gains in classification performance may be achieved by using the approach when the labeled data do not adequately represent the entire class distributions. In the second part of the chapter, we propose a quality function for learning decision boundary between data clusters from unlabeled data. The function is based on third order polynomials. The objective of the quality function is to find a place in the input sparse in data points. By maximizing the quality function, we find a decision boundary between data clusters. A superiority of the proposed quality function over the other similar functions as well as the conventional clustering algorithms tested has been observed in the experiments.