Abstract
The project focuses on the classification of the social and psychological parameters of persons in crowded places with the poor protection. This information will be extracted from a video data typical for a common surveillance infrastructure (such as a railway station or shopping center system) using methods of psychological analysis, machine vision and deep neural networks. The parameters obtained will serve, among other things, to determine the statistical distribution of the pre-evacuation time in the event of an emergency, when this time interval is for security personnel critical and hard to obtain, and generally to improve supervision. Used methods will comply with the General Data Protection Regulation.