Tracking people in RGBD videos using deep learning and motion clues

Tracking people in videos is an important topic in surveillance. We consider the problem of human tracking in RGBD videos filmed by sensors such as MS Kinect and Primesense. Our goal is to track persons where the crowd of people is known in advance or all persons in the video have appeared in the very beginning. Thus we can train a classifier to help classify and track persons across the video. A deep learning model trained with big data has been proved to be an effective classifier for various kinds of objects. We propose to train a deep convolutional neural network, which improves tracking performance, to classify people. And a motion model based on spatial and kinetic clues is combined with the network to track people in the scene. We demonstrate the effectiveness of our method by evaluating it on several datasets and comparing with traditional methods like SVM.