Sarah Deane and Matthew Sharifi from the University of Southampton conducted their studies because of the growing demand for intelligent CCTV in security and marketing applications.
Deane’s project found most CCTV footage fails to provide a clear image because it is obscured by background information. Several techniques exist for removing this background ‘noise’ but as she explains, none is particularly effective.
‘I found that background subtraction, although being simply defined as a difference between the background image without objects of interest and an observed image, has many difficult issues to overcome,’ said Deane.
‘It was apparent that a simple subtraction algorithm was needed to allow the high computational efficiency that is required by CCTV applications.’
Background subtraction uses algorithms to isolate moving objects from a wider CCTV image. Traffic, changing weather and crowded areas can all complicate the process and Deane’s report, which is now available online, concludes that fast processing is key to their success.
Sharifi’s research compared the use of facial recognition techniques with Bluetooth for identification purposes.
He found that using a CCTV camera placed on a reception desk he could detect and identify every person that walked past. In the Bluetooth experiment, only 8.33% of those involved could be identified and even this was reliant on them carrying a Bluetooth device.
The results have inspired him to conduct a much larger video dataset and further experiments.
‘Having observed the advantages and disadvantages of both Bluetooth and face recognition, it would be interesting to combine the two techniques into a multi-modal identification technology which could couple the ubiquity of face recognition with the recognition accuracy of Bluetooth,’ he said.
