Towards A Self-Calibrating Video Camera Network For Content Analysis And Forensics