Carnegie Mellon Robotics Institute
Wine drinking may not have changed much over millenniums, but the same can't be said for the winemaking process.
Chosen techniques for grape growing, fermenting and bottling of wine often distinguish winemakers, yet some technology researchers believe Moore's Law could help run vineyards of the future.
While many winemakers around the world have been experimenting with new approaches such as chemistry to fine-tune the taste of wine, computer technologies like wireless sensors are being used to control irrigation of some vineyards. Recently, researchers have been putting computer vision technologies to work in vineyards, believing that one day winemakers might even be able to use their mobile phones to actually see and help manage their crops.
Predicting a crop's yield has long been a common practice among grape growers, but a few, like Wente Vineyards in California, consider it an exact science.
Debadeepta Dey, PhD The Robotics Institute, Carnegie Mellon University is part of the research team using computer vision research technology at Cornell University's vineyard
Karl D. Wente is a fifth-generation winemaker at Wente, which claims to be the country's oldest, continuously operated family-owned winery in the country. His family blends traditional and innovative winemaking practices as they cultivate approximately 3,000 acres of sustainably farmed vineyards. Wente estimates annual yields with the naked eye, looking at the number of vines per acre and clusters per vine multiplied by an assumed weight for each cluster.
"It's actually become a running joke around the winery about how tough it is and how far off we are on our estimates," said Wente in a Future Lab podcast sponsored by Intel Labs. "We could be off 20 percent in either direction."
Wente creates wine from whatever amount of grapes he can harvest each year, but other grape growers, especially those dependent on selling their harvest to other wineries on the open market, could potentially benefit from being able to more accurately predict the amount of grapes their annual crop will yield.
"Being off 20 percent on the low side could mean that the grower might bring in 80 percent of planned income," said Intel Labs researcher Richard Beckwith, who has helped carry out technology research in agriculture areas for more than a decade, including wireless sensor technology at Oregon's Cameron Winery in 2003.
"In addition, fruit quality is a function of crop load," said Beckwith. "Many winemakers believe that you have to keep the crop load low to get the best quality. Being off by 20 percent on the high side could compromise the quality of the grapes. It is an important balancing act."
A team of researchers is putting technology to work at various vineyards around the country, including the Cornell University vineyard just off the shore of Lake Erie where Concord grapes are grown, and at Pinot Noir vineyards in Oregon.
The team, which recently applied for a Department of Agriculture grant, includes scientists from Intel Labs, Penn State and Cornell faculty, Carnegie Mellon University Robotics Institute faculty and students, and experts from the grape industry. The team hit the field last fall equipped with digital cameras, laptops and sophisticated software, and began monitoring the vigor of grape vines and figuring out how computer vision can someday help grape growers estimate harvest yields more reliably than today's histrionic approaches.
"We use shape to distinguish between leaves, grapes and linear structures, and from that we can try to figure out how much fruit there is," said Lily Mummert, an Intel research scientist who collaborated with Carnegie Mellon U, Cornell and others on recent research.
Researchers capture pictures up and down the rows of vines then use a computer program to create a three-dimensional reconstruction that could accurately distinguish between a grape, stem and a leaf, before counting each and every one.
"There's the idea of a balanced vine that has a proper balance between the energy-producing canopy leaves and the amount of fruit," she said. "Measuring both helps us get a better balance."
Members of the research team say they collected some encouraging results, but it's said it's still early, with many challenges to overcome to create a system that is accurate and reliable. Work will continue as a group of viticulturists, engineers, scientists, economists, educators and industry representatives work together on efforts such as the Vineyard Efficiency organization and file for government funding.
"This project encounters many of the standard problems you run into with computer vision," said Mummert. "You have an unstructured environment, variations in lighting, you have occlusions. So one of the challenges we have is the fruit zone occluded by leaves, and we have to figure out how to deal with it."
Several colleges in the United States are offering art and science of winemaking undergraduate, with some even offering graduate degrees, including Fresno State, Sonoma State and UC Davis in California, Cornell University in New York, Virginia Tech and Washington State University.
James Kennedy, a wine chemist at the Australian Wine Research Institute, in an interview with Wired, said that looking across the continuum of basic science to applied science to technology, science is changing wine. "The easiest place to see this is if you look at the advancement of many of the world's newer wine regions, and the rapidity with which they have become world-class producers of quality wine. Much of this can be attributed to science and technology."
While it makes sense for well-known traditional vineyards and wine-producing regions to stick with proven winemaking methods, for many the future of winemaking may be about picking the right customs and innovations.