An object tracking method based on CNN and optical flow
Luyue Lin, Bo Liu, Yanshan Xiao · 2017
Object tracking is a problem about computer vision and image understanding. It is widely used in the field of military navigation, traffic control, video surveillance and security of public life. In this paper, a new object tracking algorithm, which is based on Bayesian Filter framework, convolution neural network (CNN) and KLP Optical Flow is proposed. Our method comes up with an adaptive ingrate method to introduce a competition mechanism between CNN and KLP Optical flow and compensate the over-fitting. The experimental results show that the algorithm is able to track the target in the video stream and is robust to the challenging aspect of Motion Blur, Scale Variation and Fast Motion.