Classification of Anthurium flower cultivars based on combination of PCA, LDA and SVM classifier

Abstract

Two main steps in object recognition systems involve feature representation and classification. In this study, the combination of principal components analysis (PCA), linear discriminant analysis (LDA) and support vector machine (SVM) approaches were used to develop a cultivar classification system for Anthurium flower, which PCA, LDA, and SVM were applied for data reduction, feature extraction, and classification, respectively. The system was tested on a database of Anthurium flower images, which included the images of 20 cultivars of the flower with different sizes, angles of rotation placed under the imaging camera, and lighting conditions. Results were evaluated from the two point of view of classification accuracy and processing time, and the approach had remarkable results when trained via suitable multi-class SVM classifier features, as it is possible to increase the classification accuracy up to 99.5%. Variety recognition of flowers is an important step for subsequent flower real-time grading tasks and such algorithms could be used for these procedures.