Infrared Target Tracking Based on Robust Low-Rank Sparse Learning

Abstract

In recent years, the low-rank sparse tracker has been successfully used in object tracking by exploiting low-rank constraints to capture the underlying structure of candidate particles. It uses simple sparse error to account for occlusion and noise measured by the L1-norm, which is assumed to be following the Laplacian distribution. However, this Laplacian assumption may not be accurate to describe complex corruptions. In this letter, we propose an infrared (IR) target tracking method based on a robust low-rank sparse representation which aims to seek for the maximum-likelihood estimation solution of the residuals in the tracking framework. Experimental results on challenging IR image sequences indicate that the proposed method achieves favorable tracking performance and is more robust to various types of noise.