IP for Licence
Real Time Video Analysis
Each day in 2019, around 2,000 petabytes of video data are generated by security cameras around the world, up from 500 petabytes per day in 2015. This is equivalent to 75 million users streaming an hour’s HDTV simultaneously1. This dramatic growth is driving the demand for more efficient methods of analysing video data. New video analytics from the University of Warwick can outperform other methods, whilst requiring less processing and training time, and less storage. We are now seeking trial sites with a view to future licensing.
Dr Victor Sanchez and his team at the University of Warwick have designed a new method of feature descriptor for video analytics that encodes the motion information of a Spatio–Temporal support region into a low-dimensional Binary string (STB).
The STB descriptor is able to address these needs for classification, action recognition, object recognition, video surveillance, monitoring and abnormal event detection in video captured in real environments.
The encoded motion information is obtained from two motion sources: optical flow and temporal gradients, which provide rich motion information by considering pixel intensity changes to create a new data space that disregards the background
- low computational times and reduced memory and storage requirements
- increased descriptive power through encoded information from two motion sources
- much shorter training times compared to CNN based systems
- suitable for real-time applications and in devices with low-computational capacity
- significantly outperforms other binary descriptors including 3D-FREAK, 3D-BRISK, 3D-ORB & 3D-BRIEF
International patent application number:
PCT GB2018/058103 26 October 2018
For further information contact: Dr Shum Prakash, Warwick Ventures
Tel: +44 (0) 24 7657 4145 Email: firstname.lastname@example.org