Localising, understanding and manoeuvring
Localising and understanding object motion patterns in a scene captured by a camera is fundamental to allow autonomous decision making. The task of localisation becomes increasingly challenging as the scene gets more populated of objects to localise and less constrained, as objects' appearance gets less distinguishable, as objects' motion gets less predictable and as cameras' degrees of freedom increase. This talk will cover methods for the localisation of multiple objects in dynamic scenes and methods that use the estimated motion patterns to understand object behaviours. I will show how these estimated motion patterns can be processed by flying cameras to perform autonomous manoeuvres such as avoiding moving objects.
Speaker: Dr. Fabio Poiesi (Researcher, Fondazione Bruno Kessler)