Robotics lured Scarlett into the world of wine

That Scarlett Liu became an engineer is not really surprising, but her move into the world of wine research was a little unexpected.

Born in Changsha, the capital of Hunan province in south central China, Scarlett had science in her genes; her father worked in a university and an uncle is a Professor of Applied Physics.

Her own interest was in mechatronics and robotics, and after completing an undergraduate degree in mechanical and manufacturing engineering she moved to Sydney to start a PhD.

‘I really wanted to come UNSW Sydney because they have a robotics group within the School of Mechanical and Manufacturing Engineering that applies robotics techniques in the field, which is exciting’, she said. ‘The work is very varied; from hardware design to machine learning and data mining.’

The School also has a Smart Robotic Viticulture group led by Dr Mark Whitty, who was Scarlett’s PhD supervisor. He was looking for someone to be part of a Wine Australia-funded project investigating improved yield prediction for the wine sector. Scarlett grabbed the opportunity.

‘I think it is a very interesting sector and there is great potential to apply a range of robotics techniques’, she said. ‘It can produce many useful outcomes in the future. There are a lot of things that used to be done manually that we could now automate and we look forward to working with more partners in the Australian wine sector to increase their competitive edge.’

After four years of work, Scarlett finished her PhD last year and the now Dr Liu is currently employed as a Research Associate, working on the final stages of the project and the report that will be delivered in September.

Prompted by Angus Davidson at Treasury Wine Estates, the project set out to test whether proximal image sensing technology could be used to directly estimate fruit load and berry size at any time during the growing season. In contrast with remote sensing approaches that just assess the canopy, this approach assesses the fruit load and berry size directly.

‘I am still processing the data collected in the third year but the outcomes are looking really good right now’, Dr Liu said.

‘We can improve the yield estimate based on the shoot count that is extracted from sensor images and do this earlier than anyone else in the world.’

That is significant because the existing best practice systems used after fruit set, which are based on manual measurements, are on average 15 per cent out, which is substantially greater than the 5 per cent winemakers and wine companies would like.

The success of the new approach has already attracted interest from overseas as well as Australia, and Scarlett hopes it is just the start of a research career. ‘I’m interested in the potential of robotics generally, but the field of viticulture is definitely my priority area now. It’s really intriguing and interesting.’