Deep Fundamental Diagram Network for Real-Time Pedestrian Dynamics Analysis

The fundamental diagram of pedestrian flow, which describes a relation between pedestrian velocity and crowd density, is an important means forpedestrian dynamics analysis. Some recent work calculates the fundamental diagram of pedestrian flow by tracking each pedestrian in the crowd from video recordings. However, such methods are limited in representation of crowd density and hard to achieve a real-time analysis. To address this problem, this work proposes a novel convolutional neural network-based framework, called deep fundamental diagram network, for real-time pedestrian dynamics analysis. Our proposed framework is consisted of two parts, the multi-scale recursive convolutional neural network (MSR-Net) and an optical flow module, accounting for density distribution estimation and pedestrian motion prediction respectively. Specifically, MSR-Net is presented to learn the direct mapping from the input image of pedestrian flow to the output map of crowd density. Optical flow method is introduced to predict the velocity and direction of pedestrian in real-time. In this way, by aligning the position of pedestrian density map we are able to obtain the fundamental diagram, which shows good agreement with the ones from classical methods but higher computational efficiency. Simultaneously, deep fundamental diagram network can detect anomaly activity of pedestrian (In this work, the anomaly is defined as sudden stop and acceleration, reverse walk.), which is also meaningful for crowd analysis.