Weed Seeds Identification by Machine Vision

GRANITTO, PABLO M.; NAVONE, HUGO D.; VERDES, PABLO F.; CECCATTO, HERMENEGILDO A.

Computers and Electronics in Agriculture

Año: 2002 vol. 33 p. 91 - 104

The implementation of new methods for reliable and fast identification and classification of seeds is of major technical and economical importance in the agricultural industry. As in ocular inspection, the automatic classification of seeds should be based on knowledge of seed size, shape, color and texture. In this work, we assess the discriminating power of these characteristics for the unique identification of seeds of 57 weed species. We found that, as expected, size and shape characteristics have larger discriminating power than color and textural ones. However, all these features are required as classification parameters to reach an identification performance acceptable for practical applications. A careful selection of seed characteristics lead us to identify a nearly optimal set of 12 (6 morphological + 4 color + 2 textural) classification parameters. Using these parameters we compared two different classification strategies (naïve Bayes and artificial neural network committee) for the identification of seed species. We found that the artificial neural network committee performs slightly better than the naïve Bayes classifier, although, in spite of its simplicity, this last method is surprisingly good for the identification problem.