Managing and re-using large scale video archives

The Problem

TV and film producers already use large amounts of pre-exisiting media in new productions. As more and more media becomes available and production schedules and budgets shrink, their need to handle it efficiently is growing rapidly. Can we re-use valuable large scale wildlife video archives by retrieving relevant material in real-time and exploiting them in the creation of new programmes?

The Research

• To develop tools to create an intelligent multimedia management server.

• To automatically segment, summarise and index rushes.

• To classify specific types of species, their motion and behaviour.

• To develop individual animal recognition tools using biometrics.

• To retrieve the best shots from archive, matching researchers' and editors’ specific requirements.

• To analyse scenes invariant of their composition, camera angle or lighting conditions.

• To generate tools for fast and intuitive navigation and browsing through terabytes of digital video.

The Achievements

• Automatic shot detection, intelligent key frame extraction and camera work classification systems.  This tool uses the digital data of a shot to identify its type, camera movement and lighting etc. and then extracts one frame that best summarises the whole shot, allowing easy comparison. 
• Algorithms for gait analysis, animal tracking and identification have been developed so producers can search through archived footage for a particular animal.
• A taxonomy for wildlife video production.
• A system for individual penguin identification, implemented at Robben Island,South Africa.
• Semantic searching using a classification structure.
• Searching by a given image or shot example.
• Biometric recognition of species and their behaviour.
• Segmentation, region and motion analysis.
• The creation of a large-scale centralised multimedia server.