Distinction of “Sangiovese” clones and grapevinevarieties using Elliptic Fourier Analysis (EFA),neural networks and fractal analysis

Parameters coming from Elliptic Fourier Analysis (EFA) and fractal analysis of the leaf were used as input in a backpropagation neural network to distinguish between ten accessions of grapevine (eight clones of ‘Sangiovese’ and two varieties, ‘Abrostine’ and ‘Pugnitello’). The results enabled us to identify all the accessions studied. Cluster analysis of the backpropagation outputs revealed the existence of a uniform group for the ‘Tebano’ clones supporting the hypothesis that ‘SG-12-T’ and ‘SG-14-T’ could be mutations originated from the same genotype. The use of neural networks coupled to new tools in the shape analysis of leaves opens interesting prospects for ampelography.