Virtual Panning
for Lecture Environments

Problem

Automated lecture recordings enhance learning, but most solutions use proprietary software and expensive equipment such as Pan- Tilt-Zoom (PTZ) cameras. These solutions can cost upwards of R60,000.

This Pan Tilt Zoom (PTZ) camera is able to locate a presenter in real time, however it is expensive.

We were approached by the Centre for Innovation in Learning and Teaching (CILT) to develop a cost-effective solution to the problem of automated lecture recording. This would aid them in their research on how to reduce the cost of open courseware.

Proposed Solution

Our team designed and implemented a lecturer recording system in C++ using the OpenCV library. The system uses two to three static cameras instead of a PTZ camera. The system consists of three different components that interact with each other.

This diagram shows how the stitching, tracking and panning components fit together in the system.

The stitching component takes in multiple (overlapping) camera recordings and stitches them together to produce a panoramic video of the lecture theater. The tracking component detects the position of the lecturer within this video and stores their coordinates along with the frame number. The panning component uses the stitched video along with the coordinates to produce a cropped, lecturer-focused recording of the lecturer.