Adaptive Real-Time Compressive Tracking

The real-time compressive tracking algorithm, proposed by Kaihua Zhang etc. in 2012, is real-time and robust. But this algorithm may lead to object tracking losing in some complex environments, such as pose variation, illumination change, occlusion, and motion blur etc.. This paper improves the compressive tracking algorithm in two aspects: (1) we propose a self-adaptive method for learning parameter of the compressive tracking algorithm to enhance the robustness; (2)To solve the problem of losing tracking object, we propose a method using cosine algorithm to judge whether the object is lost and retrieve the lost object again. A number of video object tracking experiments show that the improved algorithm is more effective and efficient.