Ben Chostner 2017-06-28 00:21:29
Precision agriculture has been steadily narrowing the focus of farm management—from region to field to individual plants. Each step has required technical breakthroughs and new products that add technology to farm machinery and, more importantly, profit to farm operations. Yield monitors, GPS-based autosteer, variable-rate application controllers, and electronic seed metering are becoming standard equipment, and that’s just the beginning. The origins of Blue River Technology In 2011, two Stanford graduate students, ASABE member Jorge Heraud, the former head of precision agriculture at Trimble, and Lee Redden, a PhD student and roboticist, met while taking Steve Blank’s Lean LaunchPad course, a new class to teach engineers, scientists, and other professionals how startups really get built. As classmates, Heraud and Redden discovered that they shared an ambition: to make farming more sustainable through robotics, machine learning, and computer vision. Together, they tested their idea in California’s Central Valley to prove its applicability to agriculture. With a big precision ag idea, support from friends and family, and a grant from the NSF, Blue River Technology was born. Blue River Technology, now a company with a presence in both Silicon Valley and U.S. farmland, is advancing precision agriculture by taking farm management to individual plants. Adding cameras, computers, and artificial intelligence to farm machinery allows farmers to understand and manage every plant throughout the entire season and across every operation. This smart equipment can make every plant as productive as possible while using the fewest inputs. Ultimately, the company aspires to equip all farm equipment— including tillage machines, planters, sprayers, and harvesters—with smart technology. The company’s first smart machine—the LettuceBot—focused on lettuce thinning, which traditionally has been an expensive and labor-intensive task. The LettuceBot automated this arduous chore by using imaging to identify weeds, to control spraying, and to verify the performance of the system. The LettuceBot received a 2017 AE50 Award. The next generation See & Spray is the next generation of Blue River’s weed-control technology. This tow-behind smart sprayer uses “deep learning” algorithms to identify a greater variety of plants and weeds with better accuracy, it uses custom nozzle designs that provide spray accuracy down to 2.5 cm (1 in.), and the improved software allows faster and more agile operation. See & Spray is based on some of the same technology used in self-driving cars to allow the sprayer to see all the plants in a field, distinguish crop plants from weeds, and apply a targeted dose of herbicide only on the weeds. With smart spraying, farmers can introduce new herbicides into their management practices to control herbicide-resistant weeds while reducing herbicide use by up to 90%. “I’ve been dreaming about machines that can manage every plant,” said Heraud. “Seeing them become a reality in the field is phenomenal.” Achieving sub-inch accuracy while moving through the field at 6 to 13 kph (4 to 8 mph) required close coordination between mechanical, software, and machine learning engineers. “Taking precision electronics and robotics out into the field and making them rugged is a balancing act,” said ASABE member Tyler Niday, a mechanical engineer at Blue River. “For example, we have to design components that maintain a camera’s orientation to the crops to within a degree or two, all while taking a beating in day-to-day use. We found a lot of ways that didn’t work before finding the one that did.” See & Spray is a 9 to 18 m (30 to 60 ft) wide towed implement comprising a customized 18 cm × 18 cm (7 in. × 7 in.) toolbar that supports one spray unit per crop row. “We had to rethink every component of a standard sprayer,” noted Bryon Majusiak, mechanical team lead. “We were able to use a lot of tried and true agricultural components, like pumps, filters, and fuses, while designing our own nozzles, row units, and power system.” Hardware wasn’t the only challenge in bringing See & Spray to life. Every field, every plant, and every weed is slightly different, and the software team had to develop a crop and weed identification system that could handle that variability. “Deep learning has been a great tool for this problem,” said Rajesh Radhakrishnan, senior computer vision engineer. “Traditionally, we would have designed an algorithm by using visual cues to identify the 10 or 20 parameters that best distinguish crops from weeds. With deep learning, we can design a neural network that learns a million parameters to determine which are best for distinguishing crops from weeds. We do this by training the neural net with hundreds of thousands of examples from different farmers’ fields.” Willy Pell, Blue River’s Director of New Technology, sees this new product as only the beginning of what deep learning algorithms can do. “Deep learning has been around for many years, but the innovations in software and hardware that we’ve seen in the last couple years are incredible. We’re experimenting with mobile processing units on many pieces of farm equipment. A smart planter, smart sprayer, and smart harvester can work together to optimize the profitability of every plant in the field.” Blue River is demonstrating the See & Spray weeding machine in the U.S. cotton belt this year and will begin commercialization in 2018. The use of modular row units and deep learning software make See & Spray easily adaptable to new crops, and the team is currently training the system with additional crops like soybeans, corn, and peanuts. “Farmers have been asking for better tools to manage their farm and maintain profitably,” said Heraud. “Artificial intelligence and robotics, if packaged together into a great product, could be just the tool that farmers are looking for.” Ben Chostner, Vice President, Business Development, Blue River Technology, Sunnyvale, Calif., USA, firstname.lastname@example.org. To learn more about Blue River Technology and the See & Spray, visit www.reuters.com/video/2017/04/17/generation-maker-weeds-be-wary-these-rob?videoId=371503493&videoChannel=118245
Published by ASABE. View All Articles.