IDENTIFICATION OF, AND DISCRIMINATION AMONG, LUCERNE (MEDICAGO-SATIVAL.) VARIETIES USING SEED IMAGE-ANALYSIS

Seed image analysis has potential for discrimination among varieties o f self-pollinating species. This study teas conducted to determine whe ther this technique could be used for differentiation and verification of varieties of a cross-pollinating species, lucerne (Medicago sativa L.). A digital image processing algorithm (VIPS) was used for image p rocessing of 150 individual seeds of 17 seed lots of eight lucerne var ieties. Of the 21 morphological characters and derived measurements re corded, 10 were potentially valuable for verification of, and discrimi nation among varieties. Of these 10 characters, the ability to discrim inate at P<0.001 among all 28 possible pairings of the eight varieties ranged from 71% (seed blue colour/total colour intensity) down to 11% concavity/width. A better variety discrimination (82%) tons obtained on the basis of a combination of the 10 selected characters using Cano nical Discriminant Analysis. Comparison of the squared Mahalanobis dis tances among the varieties using the standard seed lots and these 10 s eed morphological characters produced results for relatedness among th e varieties which were generally consistent with known geographical an d breeding relationships. These results suggest that seed image analys is cart be useful for providing data both for registration of new luce rne varieties, and for lucerne seed certification.