Artificial vision and morfometry in analysis and classification of biological species

Rodrigo de Oliveira Plotze

Instituto de Ciências Matemáticas e de Computação - Universidade de São Paulo, 2010

The discovery of taxonomic and evolutionary history of species is the main frontier of scientific research in various knowledge areas. The biodiversity of living things associated with the great morphological variability, makes this task a rightful challenge. The traditional methodologies based on visual inspection are totally outdated. Nowadays, technological advances have made available to researchers a framework of tools for the study of the species. This study aims to investigate and develop computer models to perform analysis and classification of biological species from morphometric features. For this, techniques of image analysis were used to determine the variability of the species in terms of three information of interest: shape, tubular structure and texture. The capability of the methods was evaluated by follows biological species: passion fruits of genus Passiflora, eucalyptus of genus Eucalyptus and rice of genus Oryza. The experiments produced a detailed dataset about the morphometric information of the species. The results showed that the techniques of artificial vision are demonstrably important to the study of the species. The shape analysis techniques indicated the viability of these methodologies in the species classification, in special the complex networks, the wavelets transform and the multiscale fractal dimension, have achieved high rates of correct discrimination. The new methods developed for analysis of tubular structures have also show great potential in species classification, especially the multiscale fractal signature from biometrics points. The techniques developed for texture analysis also show significant results in the investigation of the species. The results obtained with the approaches indicate that the relationship biology and computing is essential for the development of efficient methods. The inference of evolutionary tree of species was an important result obtained with the morphometric data collected from form, tubular structure and texture. Statistical tests showed a correlation between the morphometric data obtained by artificial vision and molecular data of phylogenetic reconstruction. A multidisciplinary approach is the focus of the work, which is related to the areas of artificial vision, morphometry and biology. Thus, this symbiosis has resulted in important contributions to the areas involved

keywords: Artificial vision,Image analysis,Morphometry,Taxonomy