Identificación de las subespecies de Magnolia cubensis (Magnoliaceae) mediante análisis digital de imágenes de las hojas / Identification of Magnolia cubensis (Magnoliaceae) subspecies by digital analysis of leaves images
Resumen
Palabras clave: contorno, elípticos de Fourier, forma de la hoja, morfometría geométrica, Procrustes, puntos clave
Abstract: The leaves of Magnolia cubensis have continuous variations, that makes difficult the identification at level of subspecies. One of the emergent solutions in front of this difficulty is the classification by digital analysis of images, with the use of the geometric morphometric. Starting from the characterization of leaf shape of both subspecies, the present investigation has the objective to evaluate two methods of geometric morphometric: landmarks and outline for the identification of M. cubensis subsp. acunae and M. cubensis subsp. cubensis. Were scanned leaves of 20 individuals of each subspecies, 14 landmarks were placed and the outline was automatically extracted and characterized by the Elliptic Fourier coefficients of the first 10 harmonics. Generalized Procrustes Analysis was performed to the coordinated “x” and “y” matrix of the landmarks. The standardized matrix and the descriptors were reduced with an analysis of principal components. To analyze the possibility to use leaf shape to differentiate the subspecies for both methods were carried out discriminant function analysis with the eigenvalues scores. The discriminant function analysis showed an average of 78.94% of correct discrimination in the analysis of landmarks and 82.05% in outline analysis. In this study was not achieved the total discrimination of the subspecies by means of geometric morphometric. In general, the analysis with landmarks allows to visualize and to characterize better the differences in the form of the leaves, while that, the analysis of outline is more efficient in the classification.
Keywords: outline, Elliptic Fourier, leaf shape, geometric morphometric, Procrustes, landmarks.