Applying Multi Support Vector Machine for Flower Image Classification

Image classification is the significant problems of concern in image processing and image recognition. There are many methods have been proposed for solving image classification problem such as k nearest neighbor (K-NN), Bayesian Network, Adaptive boost (Adaboost), Artificial Neural Network (NN), and Support Vector Machine (SVM). The aim of this paper is to propose a novel model using multi SVMs concurrently to apply for image classification. Firstly, each image is extracted to many feature vectors. Each of feature vectors is classified into the responsive class by one SVM. Finally, all the classify results of SVM are combined to give the final result. Our proposal classification model uses many SVMs. Let it call multi_SVM. As a case study for validation the proposal model, experiment trials were done of Oxford Flower Dataset divided into three categories (lotus, rose, and daisy) has been reported and compared on RGB and HIS color spaces. Results based on the proposed model are found encouraging in term of flower image classification accuracy.