An automatic method to identify social groups of pedestrians is discussed. Inspired by social psychological principles on group behavior, this method utilizes a novel Non-recursive Motion Similarity based Clustering (NMSC) algorithm to
cluster pedestrians based on their motion similarities. Individual pedestrians are detected and tracked in video scenes. Pedestrians are automatically clustered based on their pairwise motion similarity by considering their relative distance, relative speed and relative direction of motion. Pedestrian clusters which persist for a period of time are identified as social groups.
The below figure explains the various steps in the pedestrian group identification technique.
