COMBINING GLOBAL AND INDIVIDUAL IMAGE FEATURES TO CHARACTERIZE GRANULAR PRODUCT POPULATIONS

The characterization of granular product populations using image analy sis is a difficult problem because it often requires the extraction an d combination of many different features. We propose to study in a gen eral way these problems of granular product classification, considerin g the image analysis phase, the processing of the information extracte d and the decision making, In this paper we focus rather on the decisi on system development. It is based on a hierarchical approach to the p roblem, in eluding a generalist system whose outputs are ambiguous (an approximative solution), connected to specialist systems trained to g ive non-ambiguous solutions, The inputs of the generalist system are t he components of a vector containing the most important information fo r discriminating all the decision classes, while the inputs of the spe cialist systems are those which best distinguish a given class from an other. This strategy enables us to overcome the multiclass aspect of t he problem. It is independent of the choice of the techniques to selec t the pertinent information and to take the decision, This method is a pplied in the framework of a meal classification where three types of classifier (discriminant analysis, k nearest neighbours and multilayer neural networks) are compared. (C) 1997 John Wiley & Sons, Ltd.