Leaves Shape Classification Using Curvature and Fractal Dimension

The great biodiversity of species makes the plants classification a very complex and time-consuming task. The leaf is an important characteristic of the plant and it is present independently of season or plant maturity. The most relevant information about the leaf relies on shape. Its study enables us to discriminate a large set of species and to speed up the measures extraction process, which is traditionally performed manually. This paper presents a novel approach to leaf shape identification based on curvature complexity analysis. By using the Curvature Scale Space (CSS), a curve describing the complexity of the shape is achieved. Descriptors computed from this curve are used to classify a set of leaves shapes. Results demonstrate the potential of the technique, which overcome traditional shape analysis methods found in literature.