Investigation of the Effect of Noise on Tracking Objects using Deep Learning

Nowadays, tracking objects has become one of the basic needs of security systems. Deep learning based methods has dramatically improved results in tracking objects. Meanwhile, the quality of the videos captured by camera is effective on the accuracy of the trackers. All images captured by camera inevitably contain noise. The noise is usually created due to various reasons such as the underlying media, weather condition, and camera vibrations in the wind and so on. This paper deals with the issue. In this paper, tracking objects is performed by Yolu 3 architecture in deep learning. Cycle spinning method is also employed to eliminate noise.