A Real-time Algorithm for Regional Crowd Counting Based on Survival Analysis Theory

This paper proposes a real-time algorithm for regional crowd counting based on survival analysis theory. Our algorithm can accurately calculate regional people flow in real time over a long period, which considers tourists' stopping time as a type of survival time, and discretizes its continuous distribution probability as a proportion of stopping people in a monitoring region. Compared to other crowd counting methods, our algorithm calculates regional population in real time using a mathematical statistical model, which greatly reduces the dependence on hardware devices. Experiments in two real scenarios verify that the average accuracy of our algorithm is over 91% in comparison with 60% of sensing devices.