Structured compressive sensing for robust and fast visual tracking

Abstract

The application of compressive sensing to optical sensing has received significant attention recently. In this work, we propose a structured compressive sensing based tracking algorithm for intelligent optical sensing, which exploits the random feature reduction and the structured sparse representation of the target visual appearances. The robustness of the tracker can be achieved by seeking the structured sparse solution of the compressive sensing problem. The efficiency of the tracker is improved by a random feature reduction together with the Block Orthogonal Matching Pursuit (BOMP) algorithm. We conduct experiments and show that with an appropriate random reduction of feature dimension, the proposed method can achieve a more efficient tracking without losing the robustness compared with the reference trackers.