Spatial Querying for Camera-Based Tracking Platforms

We propose a set of techniques for querying camerabased tracking data effectively. We build an infrastructure for efficient processing of different types of spatial queries, including trajectory-based similarity and pattern matching queries. To achieve this, we project the raw tracking data from all the cameras into a global coordinate system and reduce the projection noise through partitioning methods. We represent the trajectories with polynomial approximations. We also introduce a user interface where user can query the data intuitively by simply sketching trajectories in a given 2d scene model.