Camera Placement

A common goal in many vision applications is to identify and track human with distinctive visual features or “tags”. Examples range from identifying numbers on sports jersey to locating faces in crowded environment. In this project, we propose a general framework for camera placement that is optimal for this "visual tagging problem". The optimal placement is computed by capturing the vision requirements via a set of binary variables over the environmental grid points and solving it using binary integer programming.

The GUI on top allows the user to input the environment with possible obstacles (black box), computes the optimal camera placement with 12 cameras (blue arrows) and simulates its coverage (sub-image on the right with brigter grayscale indicating better coverage.) Simulation results below show the red tag on the random humanoid in blue shirt to be visible in camera 1, 2, 11 and 12. To proble further, see our publication page and demo page.