Thesis CTDIA prize

The Academic Dalcimar Casanova won the second best dissertation in the field of Artificial Intelligence. The prize was awarded by the Special Committee on Artificial Intelligence of the Brazilian Computer Society (SBC-CEIA) during the 2010 Joint Conference.

To read more visit: VII Best MSc Dissertation/PhD Thesis Contest in Artificial Intelligence

Fractal analysis of leaf-texture properties as a tool for taxonomic and identification purposes: a case study with species from Neotropical Melastomataceae (Miconieae tribe)

Davi Rodrigo Rossatto and Dalcimar Casanova and Rosana Marta Kolb and Odemir Martinez Bruno


Melastomataceae is a common and dominant family in Neotropical vegetation, with high species diversity which leads to a large variation in some morphological structures. Despite this, some species of Melastomataceae are very similar in their external leaf morphology, leading to difficulties in their identification without the presence of reproductive organs. Here we have proposed and tested a computer-aided texture-based approach used to correctly identify and distinguish leaves of some species of Melastomataceae that occur in a region of Neotropical savanna in Southeastern Brazil, also comparing it with other previously proposed approaches. The results demonstrated that our approach may clearly separate the studied species, analyzing the patterns of leaf texture (both adaxial and abaxial surfaces), and achieving better accuracy (100%) than other methods. Our work has suggested that leaf texture properties can be used as a new characteristic for identification, and as an additional source of information in taxonomic and systematic studies. As the method may be supervised by experts, it is also suitable for discrimination of species with high morphological plasticity, improving the automated discrimination task. This approach can be very useful for identification of species in the absence of reproductive material, and is a rapid and powerful tool for plant identification.