A multi-organ plant identification method using convolutional neural networks

Abstract

Most of the existing studies for plant identification focus on identifying the plants with pictures of a single organ, such as leaf or flower. The information provided by single organ is always limited and sometimes confused. A new identification method, Multi-Organ Integrated PlantNet, is proposed in this paper which combines the diverse information of multiple organs. The method consists of two stages. Firstly, we use the framework of convolutional neural networks (CNN) to make the preliminary plant decision based on single organ. Then we make the final plant decision based on multiple organs using Linear Weighted Classification and Support Vector Machine (SVM). The experiments conducted on the dataset of 100 species show that our new method improve the identification accuracy by about 15% compared with the Single Organ PlantNet.