Classification of Camellia (Theaceae) species using leaf architecture variations and pattern recognition techniques

Area of Science:

• Botany
• Computational Biology
• Machine Learning

Background:

• Leaf characters are established for Camellia (Theaceae) species classification.
• Supervised pattern recognition techniques have not been previously applied to Camellia leaf data.

Purpose of the Study:

• To assess the effectiveness of supervised pattern recognition techniques for classifying Camellia species using leaf morphological and venation characters.
• To compare the accuracy of different pattern recognition methods for Camellia identification.

Main Methods:

• Utilized leaf morphological and venation characters from 93 Camellia species across five sections.
• Applied clustering, Learning Vector Quantization neural network (LVQ-ANN), Dynamic Architecture for Artificial Neural Networks (DAN2), and C-support vector machines (SVM).
• Validated classification accuracy using training and testing datasets.

Main Results:

• Dynamic Architecture for Artificial Neural Networks (DAN2) achieved 97.92% training and 91.11% testing accuracy.
• Radial Basis Function Support Vector Machines (RBF-SVM) yielded the highest accuracy at 97.92% for training and 97.78% for testing.
• A hierarchical dendrogram confirmed existing morphological classifications of the five sections.

Conclusions:

• Leaf architecture data analysis with supervised pattern recognition is highly effective for Camellia species identification.
• DAN2 and SVM methods show excellent discriminatory power for Camellia species.
• This approach offers a robust tool for botanical classification and identification.