Title :
LectureSight - An Open Source System for Automatic Camera Control in Lecture Recordings
Author :
Wulff, Benjamin ; Fecke, Alexander
Abstract :
In this work, we present our prototypical implementation of a system for real-time scene analysis and camera control for lecture recordings. A biologically inspired scene segmentation technique is used to find moving objects in the scene. An object tracker module works on the discovered foreground regions and plug-ins enrich the data model of the tracker with further analysis results like the segmentation of body parts. A scriptable virtual director module decides about camera movements and a camera operator module cares for executing those movement commands in a smooth manner. First tests in a big lecture hall gave promising results even with a very simple director strategy. The system should serve as on open source foundation for further projects.
Keywords :
cameras; image motion analysis; image segmentation; object tracking; public domain software; video recording; LectureSight; automatic camera control; biologically inspired scene segmentation technique; body part segmentation; lecture recordings; moving object detection; object tracker module; open source system; scriptable virtual director module; Cameras; Hardware; Head; Image analysis; Prototypes; Real-time systems; Tracking; camera control; gpu; lecture recording; opencl; people tracking;
Conference_Titel :
Multimedia (ISM), 2012 IEEE International Symposium on
Conference_Location :
Irvine, CA
Print_ISBN :
978-1-4673-4370-1
DOI :
10.1109/ISM.2012.94