Identifying plant species using architectural features in leaf microscopy images

• Autor(a) principal: Florindo, Joao Batista
• Data de Publicação: 2016
• Outros Autores: Bruno, Odemir Martinez, Rossatto, Davi Rodrigo [UNESP], Kolb, Rosana Marta [UNESP], Gómez, Maria Cecilia, Landini, Gabriel
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1139/cjb-2015-0075; http://hdl.handle.net/11449/177709
• topic: Automatic species identification; Image analysis; Morphological features

This work proposes an analytical method to identify plant species based on microscopy images of the midrib cross-section of leaves. Unlike previous shape-based approaches based on the individual shape of external contours and cells, an architectural analysis is proposed, where the midrib is semi-automatically segmented and partitioned into histologically relevant structures composed of layers of cells and vascular structures. Using a sequence of morphological operations, a set of geometrical measures from the cells in each layer is extracted to produce a vector of features for species categorization. The method applied to a database containing 10 species of plants from the Brazilian flora achieved a success rate of 91.7%, outperforming other classical shape-based approaches published in the literature.