License article

On the face of it, for police it's just a matter of realism

SULLEN eyes, swarthy stubble and a menacing grimace. The threatening images of criminals made from victims' memories often grace these pages.

But facial composite software is about to become more photo-realistic in an enhanced system Victorian police are testing.

The original Identi-Kit process, which saw witnesses choose from facial features taken from police recruit photos, was computerised in the 1980s.

The system will be replaced mid-year by software developed in the crime department's identification unit, the biggest unit in the country.

Acting Senior Sergeant Cameron Tullberg said the new system would also benefit from an expanded library of facial features from offenders from other states.

''It will make a huge difference. We will be able to source [from] a bank of facial components from other states that may have a higher number of another race,'' he said.


The data bank has changed to keep in touch with a changing Australia, not only in ethnicity, but with cultural trends such as eyebrow rings.

But the ability of a victim to recall a traumatic event with accuracy remains at the centre of the process.

Some studies have shown that victims recall their attacker as looking more aggressive than they do in their arrest mug shot.

''That is partially the psychology of the offender when they are photographed: they try to look more innocent,'' Acting Senior Sergeant Tullberg said.

''We do about 120 images per month, we get about 20 per cent positive return in that an offender has been arrested as a result of the facial composite,'' he said.

''That's an interesting figure,'' said Andrew Roberts, senior law lecturer at the University of Melbourne and an expert in identification as evidence, ''because in all of the research on facial composites, the accuracy rate is very low indeed, usually under 10 per cent.''

The crime department's next step is 3D modelling, similar to that used in the gaming industry. It will allow for side-profile identification, which is not possible with one-dimensional imaging.