Leaf identification using morphology and structural decomposition

This article presents a novel method of classification of plants using their leaves. Major plant species have unique leaves which are different from each other by features as shape and texture. The primary step for the identification is detection of the edges of plant leaves which is done by Otsu´s algorithm, texture features from Gray Level Cooccurrence Matrix and binary morphology. We propose a method of structural decomposition of the edges which extracts the structural signatures and quantifies the features which are independent of leaf size and orientation. This algorithm has been tested for the images of 40 plant leaves of 10 different species and it gives an accuracy rate of 67.5%.