A multi-division convolutional neural network-based plant identification system

Area of Science:

• Botany and Computer Science
• Application of Artificial Intelligence in Agriculture

Background:

• Plant species face extinction risks due to climate change.
• Accurate plant identification is crucial for conservation and agricultural research.
• Deep learning methods have shown promise in plant image analysis.

Purpose of the Study:

• To develop an effective plant recognition system for classifying plant species.
• To leverage deep learning for enhanced plant identification accuracy.

Main Methods:

• A Multi-Division Convolutional Neural Network (MD-CNN) was designed.
• Plant images were divided into nxn pieces for feature extraction using Convolutional Neural Networks (CNN).
• Principal Component Analysis (PCA) and Support Vector Machine (SVM) were used for feature selection and classification.

Main Results:

• The MD-CNN system achieved high accuracy across eight diverse plant datasets.
• 100% accuracy was recorded for Flavia, Swedish, and Folio datasets.
• Excellent performance was observed on other datasets, with accuracies ranging from 94.38% to 99.93%.

Conclusions:

• The proposed MD-CNN based system demonstrates superior performance in plant species identification.
• This deep learning approach offers a robust solution for plant recognition tasks.
• The system contributes to the advancement of plant diversity monitoring and conservation technologies.