Study of UAV tracking based on CNN in noisy environment

Abstract

Recently, there are lots of tracking methods proposed to improve the performance of visual tracking in videos with challenging situations, such as background clutter, severe occlusion, rotation, and so on. In real unmanned aerial vehicle (UAV) based tracking systems, there are various noises occurring during video capturing, transmission, and processing. However, most existing studies pay attention to improve the robustness and accuracy of visual tracking while ignoring the performance of tracking methods on videos with noise. In this paper, we investigate the performance evaluation of existing tracking methods on videos with noise. A group of noisy UAV based tracking video datasets are constructed and used to the benchmark datasets for analysis of tracking methods. Furthermore, we propose an algorithm for robustness tracking in noisy videos. The performance of 9 tracking methods is evaluated on the proposed dataset. We provide the detailed analysis and discussion on the robustness analysis of different tracking methods on videos with different variance of noises. Our investigation shows that it is still challenging for effective tracking for existing methods on videos with noise. And our proposed method shows promising results in noisy videos.