Lincoln Agritech -- a R&D company owned by Lincoln University --says it is using artificial intelligence techniques to make early-season predictions on the grape yield a vineyard is likely to harvest, by using electronic sensors to accurately count grapes.
The leader of Lincoln Agritech’s optics and image processing team Dr Jaco Fourie said profitable wine production depended on early knowledge of the grape yield likely to be harvested each season, and grape growers and wineries spent a lot of money trying to predict their grape yield each year, hiring a large number of workers to manually count sample grape bunches.
"The sensors will capture and analyse grape bunches within individual rows, and assess the number, sizes and distribution of grape bunches," Fourie said.
"We’ll then feed this data into computer algorithms, which have been designed by the University of Canterbury, to predict grape yield at harvest time."
He said new data would be added to the system each year, continuously improving the model’s accuracy and predictive power.
The study will initially concentrate on Sauvignon Blanc grapes. The team will then look at further development to count Pinot Noir grapes.
The project is funded by the Ministry of Business, Innovation and Employment (MBIE) and NZ Winegrowers. Collaborating partners include Plant and Food Research, Lincoln University, the University of Canterbury, CSIRO (Adelaide), NZ Winegrowers and local winegrowers in the Marlborough region.