Crowd monitoring and abnormal behavior detection from visual content
As a concept, crowd monitoring is generally an umbrella term and it might include either individual human behavior analysis or extended to a group of people and estimating number of participants in a specific area. Intelligent crowd-monitoring solutions can actually supply a valuable aim to the usual measures for crowd monitoring by combining state-of-the-art computer vision techniques that excel in both processing speed and precision accuracy. The actual real-time dynamic reporting of the number of passengers or crew members present, as well as the reporting of abnormal actions they might undertake, can be expanded to a variety of scenarios in maritime environment that would otherwise necessitate a large number of people dedicated entirely to perform this monitoring activity. The application, for example, includes, but is not limited to, early reporting of stressful situations or recognizing scenarios when ship’s security personnel should intervene to prevent or save individuals.
For ISOLA, the crowd-monitoring tool developed by CERTH takes input from ship’s surveillance cameras or dedicated drones and calculates the actual number of gathered persons in critical areas. Furthermore, it provides information about certain individual or crowd activities, which are considered harmful under certain conditions. The visual content is analyzed in near-real time, in order to offer the system with a higher degree of conceptual information about numerous individuals’ abnormal behavior, in other words a crowd analysis framework. When a suspicious and violent event occurs, the crowd-monitoring tool will trigger an alert to the ISOLA system.