Crowd motion detection in video by combining CNN and integral optical flow

Abstract

The paper proposes a new approach for crowd motion detection in video by combining CNN and integral optical flow. At first, definitions of crowd motion are given, along with motion parameters that can be used to perform crowd analysis. Secondly, crowd motion features and parameters are defined. Thirdly, an algorithm of crowd behavior analysis using CNN and integral optical flow is proposed. Experimental results show that, with the help of CNN, optical flow can be calculated more accurately and quickly, and by using integral optical flow, the algorithm demonstrates stronger robustness to noise and the ability to get more accurate boundaries of moving objects.