Exploring the Use of Leaf Shape Frequencies for Plant Classification Ana Carolina Quint˜ ao Siravenha Federal University of Para Vale Institute of Technology Belem, Brazil Email: siravenha@ufpa.br Schubert R. Carvalho Vale Institute of Technology Belem, Brazil Email: schubert.carvalho@itv.org Fig. 1. Leaf shape modeling in the frequency domain. This figure shows the digital image process in combination with the Fourier descriptor for extracting a leaf shape signature. From left to right: original image; binary representation; extracted leaf contour; edge-centroid distance measure; normalised Fourier coefficients. Abstract—Plant identification and classification play an impor- tant role in ecology, but the manual process is cumbersome even for experimented taxonomists. Technological advances allows the development of strategies to make these tasks easily and faster. In this context, this paper describes a methodology for plant identification and classification based on leaf shapes, that explores the discriminative power of the contour-centroid distance in the Fourier frequency domain in which some invariance (e.g. rotation and scale) are guaranteed. In addition, it is also investigated the influence of feature selection techniques regarding classification accuracy. Our results show that by combining a set of features vectors - in the principal components space - and a feedforward neural network, an accuracy of 97.45% was achieved. Keywords-Plants classification; Shape features; Fourier trans- form; Feature selection I. I NTRODUCTION In tropical regions, among all the organs that are used in plant identification, leafs are the most used ones. Mainly be- cause they are generally present when compared to flower and fruits. For example, taxonomists can rely on leaves to search for patterns that can be used to identify a plant specie. To do so, they often analyze leaf patterns such as venations and/or shape. However, the manual process of plant identification is highly dependent on expert knowledge. In addition, it is notable the shortage of expert taxonomists, which increases the demand for new tools capable to recognize plants, for example, from images, which can be useful in field tasks [1]. In this work, the identification of an unknown plant from a given leaf database involves leaf shape modeling and classification. To give an example of how a plant specimen is represent from its leaf shape consider Figure 1. In this figure, five steps are performed until a leaf signature describing a plant specie is extracted and represented in the Fourier domain. Contributions: This paper presents a methodology for plant identification and classification that combines the power of shape descriptors, feature selectors techniques and clas- sification models. In our framework, shape descriptors are rotation, translation, scale and start point invariant by using the normalized Fourier transform (FT) applied over the edge- centroid distance signatures. Feature selectors are used for both reducing the data dimensionality and increasing classification accuracy. From this study, we observed several interesting results that validates and encourages the approach used in this work. A. Related work The literature presents many approaches directed to pattern identification for image classification and retrieving. Appli- cations for botanical purposes have been taking advantage of these technological advances [2], [3], [4], [5]. Plants identifica- tion and classification, 3D reconstruction of leaves, or species characterization were some tasks that could be tested in real applications, through the years. Each plant organ gives informations to identify and to characterize species. In particular, flowers, barks and leaves are the most common structures used to obtain the most representative features from a specimen. The approximately two-dimensional aspect of leaves, in contrast with flowers and fruits, makes those eligible to a plant identifying system based on pictures. Further, in most of living tropical plants, flower 2015 28th SIBGRAPI Conference on Graphics, Patterns and Images 1530-1834/15 $31.00 © 2015 IEEE DOI 10.1109/SIBGRAPI.2015.36 297