A Hybrid Optimization Method for Image Classification with Gravitational Search Algorithm

摘要

For image classification systems, choosing the relevant image features and the parameters of classifier are the most important factors. However, previous systems focused on only one factor. The accuracy of image classification can be further improved if we consider these two factors together. The proposed algorithm combined gravitational search algorithm in feature extraction and almost simultaneously selecting the relevant features by quantum-binary gravitational search algorithm. It incorporates support vector machine to find a hyper-plane that yields the minimum number of errors in the process, keeping the constraint violation very low. To ascertain the efficiency of the algorithm, a process of experiment was conducted on two datasets.