A facial recognition technology expert says he's "extremely confident" that two brazen jewellery heists in Auckland were committed by the same person following analysis of images of the suspected thieves-who bear a striking resemblance- from security footage.

However a second expert was less certain, saying it was difficult to be sure because of the poor quality of the images.

Michael Hill jewellery store was robbed of a $60,000 diamond ring from its Westfield mall store in Henderson in April. The shopkeeper had handed the thief a white gold solitaire-cut ring, who then ran from the store.

This month a jewellery store in central Auckland was struck in the same way. A man was handed a diamond ring worth more than $36,000 and he ran from the Elliot St store with it.

Advertisement

Police released images of the alleged offenders which revealed a startling resemblance- Caucasian, dark hair, glasses, the same nose and facial hair and receding hairline. They were also both wearing black shoes and t shirts.

The Herald on Sunday asked two facial recognition technology businesses to analyse the images to see if the two individuals could be the same person.

Police are seeking help to identify a man who is alleged to have stolen a diamond ring from a jewellery store in Elliott Street.
Police are seeking help to identify a man who is alleged to have stolen a diamond ring from a jewellery store in Elliott Street.

Australia's Imagus Technology sales manager Fraser Larcombe told the Herald his team was "extremely confident" the images were of the same person, estimating a 65 per cent chance it was the same man.

The images were grainy and analysis of further images might reveal a better result but Larcombe said: "In this case if I was a policeman with this information, I would be looking for one individual, not two."

Cognitec Asia Pacific vice president Terry Hartmann wasn't so sure. While there was 'contextual' similarities between the pictures, including the glasses, the clothing and the parting of the hair, he was unsure if it was the same person.

He put the odds at a 50 per cent chance, but said the quality of the images weren't able to build a clear picture. "If you look at these images there's some doubt there when you look at the bottom of the earlobe," he said.

"Based on the images and the image quality, the facial recognition technology is not confident it's the same person."

Facial recognition technology is widely used throughout the world both to catch criminals, and identify victims. Last year Interpol launched a global database of facial images in order to identify fugitives and missing persons and verify mugshots.

Advertisement

But it's not just used for crime.

Facebook has also just widened its use of the technology to prevent people from making fake profiles or using other people's images without their permission.

Hartmann said casinos and banks were known to use the technology to identify important customers.

Auckland city police's Sergeant Brad Kirtlan said police believed there was a "high possibility" the man in the CCTV footage was the same person.

"We are following strong lines of inquiry regarding these incidents," he said.

"For operational reasons we are not able to discuss the technology used regarding facial recognition."

Anyone who recognises the man in both images should contact Auckland police on 09 302 6588. Information can also be provided anonymously to Crimestoppers on 0800 555 111.

View the rest of the security images here.

HOW DOES THE TECH WORK?

• Imagus loaded three images of suspect one from the Henderson heist into a database, and one image of suspect two from the Elliot St theft. The database contains 6942 images of different people. The computer is then instructed to find which image in the database looks most like suspect two. "The software came back with the top result being suspect one," Fraser said. A side by side analysis of the photos came up with the 65 per cent likelihood.

• Cognitec Systems develop a 'match score' of the images in a similar way by comparing them alongside each other and against others.