Leaves classification using neural network based on ensemble features

Abstract

An automated plant identification is necessary to identify plants, especially rarely seen ones. In this paper a framework to identify plant species based on leaf's characteristics is introduced. First, 31 features of leaves from 13 species are extracted that represents color, shape and texture of the leaves. Then, the features are selected according to their correlation to the class label. The data with 25.8% pruned features are then used to train a feedforward neural network. The network is trained and tested using 975 images by implementing 10-fold mechanism yields 95.54% accuracy.