Identifying Sources and Sinks and Detecting Dominant Motion Patterns in Crowds

The construction of origin-destination matrixes is a necessary activity for most significant applications of pedestrian simulation and it can be fruitfully supported by computer vision techniques. Pedestrians, in videos taken from fixed cameras, tend to appear and disappear at precise locations (doors, gateways or edges of the scene): we refer to locations where pedestrians appear as sources (potential origins) and the locations where they disappear as sinks (potential destinations). In this paper we propose an original technique to identify these points and characterize dominant pedestrian flows. The paper presents the defined technique and it discusses its application in a real-world scenario