Metrics on Crowd Control with Overhead Video and Vocal Commands

This thesis presents an agent-tracking framework for semi-structured, crowded video. This framework is used to investigate how large numbers of people respond to vocal commands with local feedback and an overhead camera video. We analyze a video showing an overhead view of more than 200 people, each holding an umbrella equipped with red, blue, and green LED lights. The crowd’s motion under the vocal command formed a variety of patterns. We use k-means clustering to separate umbrella from each other. Kalman filtering is used to estimate how each umbrella moves and track their motion path. In particular, we present results on: (1) Automatic segmentation and classification of each umbrella. (2) Swarm’s response time to a simple command. (3) Time constant for a harder command. (4) Comparing accuracy. (5) “Shape-matching” ability. (6) Documenting the position memory. (7) Distribution consensus simulation.