Mitsubishi develops real-time data analysis system for crowd safety
Mitsubishi Electric has developed a new technology that could help avert accidents at crowded events. The system analyzes data from surveillance cameras to provide real-time estimates and predictions of crowd congestions on pathways leading in and out of venues. The technology was developed in collaboration with the Research Center for Advanced Science and Technology of Tokyo University (RCAST).
The system calculates flow rates in real time by analyzing motion information from the video streams of each of the surveillance cameras at an event. RCAST's high-speed human behavior model simulates crowd movements in a way that shifts the focus away from the overall crowd (which is what conventional methods do) to individuals and the people nearby.
The crowd-movement simulator predicts crowd congestion by working with data points including the people flow rate, pathway information and human behavior models.
The predictions of crowd congestion are then presented via a heat map over a map of the area and serves as something of an early-warning system. Armed with the predicted hot spots, a security company can promptly decide and take appropriate action to flow foot traffic in a different way to relieve areas of congestion.
Mitsubishi says the system is the world's first of its kind to simulate crowd movements in real time, with an accuracy rate of 80 percent, up 30 percent from current accuracy rates.
Mitsubishi Electric carried out a field demonstration of its new system at the 38th Tamagawa Fireworks Festival in Tokyo, which took place last weekend.