APPLICATION OF FUZZY C-MEANS CLUSTERING FOR SEED DISCRIMINATION BY ARTIFICIAL VISION

The Fuzzy C-Means Algorithm (FCMA) was applied for the discrimination between seed species by artificial vision. Colour images of seeds belo nging to 4 species were acquired with a CCD camera. In order to charac terise the morphology of the seeds, a set of quantitative features wer e extracted from the images. The aim of this study was to cluster, by FCMA, the measured learning and test data into 4 groups. The FCMA proc ess was improved by the introduction of a non-random initialisation of the cluster centres. In addition, the Mahalanobis distance was used, instead of the Euclidean distance, as a measure of the proximity of a pattern to a cluster. Furthermore, a classification approach with a re ject option was investigated for increasing the classification perform ances. This was achieved by assigning the seeds which were lying in th e fuzzy boundaries between the available classes to a reject class. Th e proposed initialisation method outperformed the random initialisatio n both in terms of the computation time and the percentage of correct recognition. With the Mahalanobis distance, the error of classificatio n was 5.32% and 6% for the training and test sets. In comparison to th e Euclidean distance, the Mahalanobis distance allowed a decrease of t he classification errors by 1.12% and 1% for the training and test set s, respectively. The main advantage of the Mahalanobis distance is tha t it takes into account the real underlying shapes of the clusters. Th e application of FCMA with a reject option showed that the classificat ion errors dropped to 3.38% and 5% when 7.43% of the learning seeds we re rejected. (C) 1997 Elsevier Science B.V.