An Enhanced TLD Algorithm Based on Sparse Representation

Abstract

To build an efficient data processing module in TLD tracking algorithm, the samples of foreground targets and the background were compressed by using a very sparse measurement that can extract the features by a non-adaptive random projections efficiently, and after the sparse representation, the dimensionality reduction data can preserve most of the salient information and allow almost perfect reconstruction of the signal. Building a real-time long-term tracking system based on the sparse representation could improve the efficiency of tracking algorithm, thereby solving the problem of efficiency decline in TLD with the time going. In our algorithm, the sparse representation combines with the three sub-tasks of tracking task: tracking, learning and detection, which can not only guarantee the ability of estimating errors, but also improve the efficiency of data processing.