Digital Tracking Algorithm for Real-time or Recorded Video Sequences - Isis Project No 4335

A method of tracking identifiable entities in video, using probabilistic methods to maintain continuous tracks of the entities, that works even when multiple entities cross paths.

Multiple Target Tracking

The digital tracking algorithm is a novel application of the ‘multiple target tracking’ theory, from defence to the concept of tracking entities on live video camera feeds or stored video. Using probabilistic concepts to link entities from frame to frame, the technology uses the past path of an entity to make a probabilistic prediction about where an entity will be in the next frame. This allows maintenance of continuous tracks even when multiple entities cross paths. The method is not limited by frame rate or time, ensuring complete flexibility of implementation.

Easy Identification and Improved Information

The algorithm can be implemented on any video system, live or recorded, where individual entities can be identified. This can range from bacterial cells under microscope camera, to humans in crowds. The algorithm has been implemented on a real life system. As well as providing visual tracks, information on speed and direction of an entity, change of direction and other parameters are all calculated and stored.

Patent Status

The Oxford invention is the subject of a patent application. Isis would like to talk to companies interested in developing the commercial opportunity. Please contact the Isis Project Manager.

Request Further Information: Project Number 4335 - Digital Tracking Algorithm for Real-time or Recorded Sequences