Abstract:

The recognition of plants species based on leaf characterization is one of the first approaches used for the design of plants taxonomy. The shape is the most used characteristic of plant leaf, it is considered to be a stable feature when used for the recognition of plant species, using leaf image. Convolutional Neural Network(CNN) design is inspired from the human brain, it has been used for object recognition with success. This paper explores the usage of CNN to improve the recognition process of plant species using leaf images. To explore the use of CNN for plant recognition, a model based on the combination of various implementations of CNN with other classifiers such as logistic regression classifier is proposed. The CNN in the proposed model is used for features extraction and the selected classifier is used for recognition. In order to evaluate this model, two experiments were conducted using FLAVIA and LeafSnap leaf image datasets. The proposed model achieves 95% with LeafSnap and 99% with FLAVIA using Vgg16+LogisticRegression