Due to crowd congestion in overcrowded areas, severe conditions such as panic and stampede may occur. These events should be predicted at an earlier stage and thus the crowd will be better managed and safe. In this paper a camera based system is introduced that predicts the congestion by analyzing the motions of the people in conflict areas as well as how their flow can become free-flowing. This system includes video, and it calculates the direction and velocity of every individual motion (optical flow) and clusters the similar motion vectors to discover key flow patterns. When counter flows come together, the system identifies such points as the conflict zones and monitors their dynamics over the passage of time to assess the risks of congestion. It also provides a turbulence index in order to quantify motion field instability. These features are fed in three prediction models, a simple linear model, a Kalman filter, and an LSTM neural network to approximate the congestion time before it happens. It is also a method used in measuring how early it can predict congestion in advance of it taking place. Real crowd video experiments indicate that incorporation of awareness of turbulence can provide reliable early congestion indicator and this method is more effective than purely basic methods.
The early detection of Alzheimer's Disease (AD) is important to slow the disease's progression and m...
The fast implementation of online examination systems has yielded a dire requirement of proctoring s...
In this paper, an algorithm is defined as an ordered set of processes or rules designed to achieve s...
This paper deals with the design and parametric study of a miniaturized elliptical slot microstrip a...
