page 1  (7 pages)
2to next section

Copyright ? 1995 by T. RayChaudhuri, J. Yeh, L. Hamey, S. Sung and T. Westcott

Appears in Proc. 8th Australian Joint Conf. Artificial Intelligence, 1995, pp. 435-441.

A Connectionist Approach to Quality

Assessment of Food Products

Tirthankar RayChaudhuriy Jeffrey C.H. Yehy Leonard G.C. Hameyy

Samuel K.Y. Sungy and Tas Westcottz

yDepartment of Computing, School of MPCE, Macquarie University

New South Wales 2109, Australia

zArnott's Biscuits Ltd., Homebush, New South Wales 2140, Australia


Colour development of a product is often vital in the food industry. The study of the baking of biscuits reveals interesting colour development characteristic curves. Neural network methods are used to both represent and classify products according to these characteristics. Using self-organising maps well-defined characteristic curves are extracted. Colour data histogrammed along these curves are then accurately classified by feedforward neural networks trained by backpropagation. Image segmentation is implicit within this colour histogramming technique. The overall system has been shown to considerably outperform a human expert.


Image Analysis, Neural Networks, Self-Organising Maps, Gaussian Filtering, Backpropagation.

1 Introduction

An important criterion for the assessment of quality of a food product is its colour. Skilled human judgement of food product quality by inspection of its colour is liable to both short-term and long-term inconsistencies. On the other hand an intelligent machine inspection system, properly calibrated and set up, is immune to such variations. The representation, analysis and classification of product colour by means of computer vision techniques is therefore becoming increasingly popular in the food industry [8, 9].

In this work we present a method of both representing and classifying the overall colour characteristics of baked biscuits | from a connectionist perspective. Our techniques are based upon the application of Kohonen's self-organising map [3, 5, 6] and the well-established feedforward neural network using backpropagation [7] (see Figure 1).