A team of engineers in Brazil has invented a machine to grade coffee beans with electronic imaging, speeding up the time-consuming task and providing information farmers can use to improve produce.

Cooperatives and exporters routinely inspect samples of beans before committing to buy, determining the grade and corresponding price by counting defects such as broken, poorly formed or discolored beans, and the presence of debris.

Classifying each 300 gram sample of coffee with the human eye takes about 15 minutes depending on quality, but inventors Helder Knidel and Leandro de Castro from startup technology firm NatComp, hope the machine, will do the job in five.

That would help ease backlogs of samples that can pile up on buyers' sampling tables at busy times of year.

"We are getting really good feedback from the industry and exporters," said Knidel, a computer engineer overseeing a team of six researchers and developers. He said they hoped to have a marketable prototype ready in six months.

Fapesp, Brazil's research promotion foundation, has provided financial backing for the project with a grant and the team, partners in startup technology firm NatComp, has also applied for national and international patents.

Knidel says the machine, about the size of a computer printer, is 96 percent accurate in detection of defects compared to an average 95 percent for the professional classifier's human eye.

As well as inspecting samples, classifiers also do skilled taste tests known as cupping, a task machines cannot perform.

"(Classification) takes time and is tiring on the eyes," said Helder, one of the participants at Encafe, a Brazilian coffee industry congress that concluded on the weekend.

Sampling by machine could also help to build trust between buyers and producers who occasionally disagree over the grade attributed to a coffee, which is used to set its price.

"Human classification is subjective. At times, you have to bring in arbiters to do the classification. With a machine you don't have this ... It will have the same result every time it looks at the sample," Knidel said.

The machine will also provide valuable feedback to the grower, detailing the type and frequency of defects -- information that classifiers do not usually provide because of the extra time it takes.

"The producer will have more information to work on improving or correcting his processing," Knidel said. Besides the climate, coffee producers say care taken during and after harvesting has the most important influence on quality.

The developers have completed software to classify arabica type coffee, the kind Brazil mostly grows. Software adapted to robusta, a coffee used more in instant coffees and whose beans are smaller, is still under development.

Knidel did not have an estimate of how much the machine would cost but expected it to be in reach of small buyers. – Reuters