3D mugshots could catch criminals

Monday, 17 April, 2000, 09:47 GMT 10:47 UK

Criminals could soon be caught with the help of a computer program that creates three-dimensional mugshots from security camera footage.

Closed circuit television (CCTV) cameras are increasingly being used to help bring criminals to justice. But the images they produce are often too fuzzy and indistinct to be of any help.

New software in development at Staffordshire University, UK, combines a series of CCTV images filmed at different angles. These are "stitched" together to produce a 3D mugshot.

Once the image is created it can be increased in size and rotated, providing an enhanced and more accurate image of the suspect.

Police tool

The mugshot program was devised by PhD graduate student James Robinson, who told BBC News Online: "My work will probably provide a starting point for a future commercially viable product. The technologies involved could be a great aid to identifying suspects on CCTV and I may get in contact with a number of police forces in the future."

While CCTV often records offences taking place it does not always reveal the identity of those caught on film.

 

CCTV cameras are placed around one location

"The combination of poor lighting, small images with low resolution and footage recorded on video tape, which is subject to distortion and noise, means generating clear and identifiable profiles is often not possible," explained Mr Robinson.

The new processing system first imports a selection of poor quality images into a computer as a series of still frames taken from different angles.

 

A mugshot is a composite of stills

The operator marks key features and characteristics of the face by plotting a series of markers on to the images.

An electronic "mesh" frame is then generated by the software and moulded onto the images, linking the highlighted key features and mapping the textures of the face.

 

A "mesh " is superimposed to get the 3D effect

Finally the program gathers common features such as the eyes, nose, mouth and hairline from the range of "mesh" models and joins them together to generate a 3D model of the face and head.

When fully developed, the process could remove the need for human monitoring of security camera footage.

In October 1998, Newham Council in London, began a trial of CCTV technology which hoped to automatically recognise the faces of known criminals and alert police.

 TOP