POC: Paphiopedilum Orchid Classifier

Abstract

Paphiopedilum Orchid Flowers (POF) are colorful wildflowers and also endangered plants since they bloom only one time per year. There are many species with a similar appearance, which makes it difficult and laborious to classify. Thus, we propose a novel Paphiopedilum Orchid Classifier (POC) based on Neural Network, utilizing the Color and Segmentation-based Fractal Texture Analysis (SFTA) features. In the classification of 11 POF species, POC achieved 97.64% of 10-fold cross validation accuracy. Besides, we also propose a new POF dataset consisting of 100 samples for each species and illustrated the prediction performance of several renowned classifiers such as Naïve Bayes, K-nearest and Decision Tree. According to research result, we hope that POC can assists botanists to classify POF for further breed selection and adaptation.