An automated potato blemish recognition system could help growers, processors and packers with quality control assessments in the future.

The system could be set up to identify various potato skin finish problems from just a few initial training tubers, Tom Duckett, director of the Centre for Vision and Robotics Research at the University of Lincoln, explained at the event.

A human expert marks up areas of blemishes and non-blemishes on the training tubers, after which the system can classify individual pixels as blemishes with high accuracy. Different varieties would need individual training, although a pre-stored training library for popular varieties could be built in, he suggested.

Once set up, the system will analyse potato tubers for up to six different blemishes, recording an infection level for each.

“Initially we see it being used for quality control assessments, particularly for potatoes within the pre-pack market,” said Adrian Cunnington, head of Sutton Bridge Crop Storage Research centre. But there was potential for a system to be developed to allow automatic grading.

Read more from British Potato 2011 on our dedicated page.