Robust visual tracking based on extreme learning machine with multiple kernels features fusion

Abstract

Extreme learning machine (ELM) has attracted attentions in machine learning fields due to its remarkable advantages such as fast operation, simple solution, and strong generalization. In this paper, a new ELM based multiple kernels features fusion method is proposed for visual tracking. Visual tracking using multiple features has been proved as a robust approach because features could complement each other. In order to capture the non-linear relationship of features, we extend the ELM into multiple kernels framework. The shape and texture features are fused into Gaussian kernel space. Experimental results on publicly available videos show that the proposed method outperforms the state-of-the-art trackers.