Video-based semantic analysis for on crowded rail stations

Video-based semantic analysis on crowded rail stations
In this project, we propose the use of advance machine learning and artificial intelligence for the semantic analysis of crowds in train stations, monitored
through a large set of non-overlapping cameras. Specifically, we will make use of deep learning neural networks and tracking algorithms for assessing and
monitoring crowd density and dynamics within train stations. Then, we propose an evidential reasoning network to extract high-level semantic knowledge on
the previous data analytics so event reasoning can be performed effectively and false positives can be filtered. This system will deliver early-warning alerts to
operators relating to: crowd behaviour, abandoned objects, loitering and crowd avoidance. The project builds on significant existing capabilities at Queen’s
University Belfast and BAE Systems Applied Intelligence Laboratories.