Jump to page content

Purdue University Engineering Frontiers

Data Science in Engineering

Remote Sensing, Better Use of Data Promise to Improve Agriculture

by Emil Venere

Purdue research is harnessing data science, artificial intelligence and remote sensing to improve sorghum varieties for biofuels by linking specific plant phenotypes, or characteristics, to their corresponding genetic information (genotypes).

The immediate goal of the project is to use advanced sensing coupled with machine-learning techniques and biophysical modeling for high throughput phenotyping, with the longer-term goal of actually connecting the phenotypes to specific genes.

The project is a partnership led by researchers in Purdue’s Colleges of Agriculture and Engineering, the Purdue Polytechnic Institute, IBM Research, and The University of Queensland. It is funded with a three-year, $6.5 million grant from the U.S. Department of Energy’s Advanced Research Projects Agency for projects focused on accelerating energy crop development for the production of renewable transportation fuels.

Mitch Tuinstra, professor of plant breeding and genetics in the Department of Agronomy and the Wickersham Chair of Excellence Agricultural Research, is the principal investigator of the TERRA project. Melba Crawford, professor of agronomy, civil and electrical and computer engineering, and the Chair of Excellence in Earth Observation, is the project lead for Engineering.

Sorghum, which is heat- and drought-tolerant and does not compete with food crops, is attractive for biofuels both in the United States and internationally. The plants can grow as tall as 5 meters and resemble corn in the early stages, but ultimately develop seed-laden panicles instead of ears.

The team is developing an automated, high-throughput system of airborne and ground-based mobile sensors to provide detailed, precise measurements of plant characteristics related to growth, development, and water and heat tolerance. Terabytes of remote sensing data are collected over agricultural plots at the Purdue Agronomy Farm, and the crops are tracked over the growing season. The data, coupled with advanced data analytics and biotechnology, will be used for gene identification and breeding — specifically for bioenergy traits suitable for use in transportation fuel.

The data acquisition and analysis platform being developed for this project can readily be extended to other types of crops and for other purposes, such as detecting disease in plants or for applications in forestry.

“The idea of collecting and analyzing this kind of data, even though the algorithms developed for soybeans or wheat might be somewhat different, offers many potential applications,” says Crawford. Being able to use remote sensing technology promises to dramatically speed up phenotyping, which is now accomplished manually in the field or via destructive sampling. Example traits include plant count and location, height, number of leaves, stalk diameter, moisture content, flowering time, chemical composition, and ultimately biomass.

“Not only is this a really time-consuming, expensive operation, but you also don’t get much data because you can only manually measure and test so many plants,” Crawford says. “So, that’s where remote sensing comes in.”

Sensors, including high-resolution RGB, hyperspectral, and thermal cameras and LiDAR, are mounted on UAV platforms and on a modified crop sprayer (referred to by the team as the PhenoRover). The PhenoRover obtains data at ~75mm spatial resolution, while the sensor-specific data from the aerial vehicles range in spatial resolution from .5 to 4 cm. The data are all precisely geolocated, allowing individual plants to be monitored through the growth season by the multi-sensor system. Automated algorithms have been developed to extract rows within plots; count and locate plants, leaves, and panicles; determine plant heights; and predict end-of-season biomass. A visualization system is also being developed to represent and display the data for interpretation.

Also leading the project are Clifford Weil, professor of agronomy; Edward Delp, the Charles William Harrison Distinguished Professor of Electrical and Computer Engineering; David Ebert, Silicon Valley Professor of Electrical and Computer Engineering; Ayman Habib, the Thomas A. Page Professor of Civil Engineering; Keith Cherkauer, associate professor of agricultural and biological engineering; Michael Leasure, associate professor of aviation and transportation technology; and Larry Biehl, systems manager at Information Technology at Purdue.

The research benefits from a range of previous Purdue investments and initiatives, including the “Purdue Move” in Plant Science that the University introduced in 2013 to broaden Purdue’s global impact, expand research, and enhance educational opportunities for its students.

Feeding the World

Another interdisciplinary agriculture-related computation and data-centric project at Purdue strives to create an open source framework and community for sharing data and algorithms. Ultimately, the goal is to improve sustainable food and agricultural systems to help feed the growing world population.

The work is led by a core of 13 Purdue researchers from diverse backgrounds, academic units and research interests. The team is organized into three sub-teams: engineering, software and computing, and food and agronomy.

The project kicks off a new Open Agricultural Technology and Systems (OATS) Center at Purdue, says James V. Krogmeier, professor of electrical and computer engineering and principal investigator of the project. The three-year, $2 million project, which began in January 2018, is funded by the Foundation for Food and Agriculture Research with cost sharing contributed by Purdue and industry.

Providing open source data could boost promising avenues for sustainable food and agricultural systems involving advanced applications, including sensing, networking, big data science, visualization and analytics. Researchers will work to improve the performance of networks linking crop production machinery through better use of data. While large amounts of agricultural data are being collected, farm machinery cannot seamlessly share data for decision-making or research at a watershed level.

“The grand challenge is to leverage these thousands of terabytes of practically inaccessible data to feed 9.7 billion people by improving crop productivity and farmer profitability while using less land, water and energy,” Krogmeier says. “Our goal is to accelerate innovation for sustainability by bringing the power of collaborative open source culture and tools to agriculture.”

The OATS Center is expected to catalyze discussion of open source software, identify and address perceived barriers, and provide pathways for collaborators.

“Adoption of open source research and development in the food-agriculture industry will dramatically improve the speed of innovation and translation to practice for innovations, just as it has done for the larger Internet economy,” Krogmeier says. “It was the proliferation of the Internet as an inherently collaborative data-sharing platform that drove open source in the technology sector. We propose that there now exists a unique opportunity for a focused effort on data and algorithm sharing in agriculture that can produce similar open source cultural shifts.”

This transformation will require both a plan for social change within the “innovation culture” of the ag industry and the technical innovations enabling such change.

“Currently, innovation in the ag industry is locked up inside the IP of individual companies, which tends to create ‘ponds’ of innovation, rather than ‘seas’ where collaboration might happen,” Krogmeier says. “So, we are attempting to get ag — especially the data science, analytics, IoT side of it — to operate a little more like Silicon Valley startups. Companies collaborate in a completely distributed way, and without antitrust concerns, by working together on an open source technology base, which they later specialize for their own IP-protected projects. This is the model of the Internet.”

The researchers also will work to improve the logistics of food systems and the supply chain, and upgrade digital equipment used in optimizing soil conditions in farming, including “integrated nitrogen management” practices to improve fertilizer efficiency. They also will work to develop open source libraries of code to advance applications of remote sensing in farming.

“We will use a blend of open source community building and educational resource generation wrapped around solid demonstrations of data exchange situations that have broad applicability,” he says. “By involving industry partners and researchers in real-time exchange projects in an open source culture, we will build community and momentum for scalable, distributed agricultural data processing that is readily translated to practice.”

The team will work with Purdue Extension to broaden the impact of the research.

“It is our belief that future University knowledge extension will depend upon thriving open source, maker-style communities,” Krogmeier says. “In this sense, University Extension will evolve into an equal partnership between University researchers, students, extension specialists and rural stakeholders.”

The project is led by Krogmeier; Aaron C. Ault, ECE senior research engineer; Larry L. Biehl, systems manager; Dennis R. Buckmaster, professor of agricultural and biological engineering; James J. Camberato, professor of agronomy; Melba Crawford, Chair of Excellence for Earth Observation and professor of agronomy, civil engineering, and electrical and computer engineering; Amanda J. Deering, clinical assistant professor of food science; Richard Grant, professor of agro-micrometeorology; Haley F. Oliver, associate professor of food science; Amy R. Reibman, professor of electrical and computer engineering; Dharmendra Saraswat, associate professor of agricultural and biological engineering; Mark A. Tucker, professor of agricultural sciences education and communication; and Mark D. Ward, associate professor of statistics. The effort will directly involve dozens of graduate and undergraduate students, facilitating an improved workforce.

Banner Photo Caption

PhenoRover is a mobile, ground-based platform that carries a sensor package capable of measuring numerous plant traits in a large number of research plots in a single day. The team also uses unmanned aerial vehicles (UAVs) equipped with advanced sensors configured to optimize the collection of diverse phenotypic data and complement the data collected from the PhenoRover.

Related Stories

Tagged as Digital Agriculture