Target tracking algorithm based on extreme learning machine and multiple kernel boosting learning

How to construct a robust classifier is always a hot research spot in target tracking based on discriminant. In recent years, multiple kernel learning combining multiple classifier to achieve better classification performance has attract wide attention. Traditional multiple kernel learning cannot be directly used in target tracking because it has a very complicated optimal question. A multiple kernel learning is proposed which is based on boosting framework. It assures target tracking can keep efficient and accurate in complicated scenes. In order to decrease the computation and increase the classify performance, extreme learning machine (ELM) is used as the base classifier. ELM has a very simple structure and rapid training speed. Compared to support vector machine, ELM has a better generalization ability. Finally, the proposed algorithm is compared with other state-of-art tracking algorithms in some challenge videos to verify the effectiveness of the proposed algorithm. © 2017, Editorial Office of Systems Engineering and Electronics. All right reserved.