The Benchbot research project is a collaborative effort at NC State between the Department of Electrical and Computer Engineering and the Department of Crop and Soil Sciences.
The idea of the project is to build a robot with machine learning that can automate the process of plant imaging. This monitoring of plant phenotyping helps researchers collect data regarding how big they grow, leaf shapes and sizes, and other physical characteristics of the plants.
From this collection of data, scientists can learn more about plant health and structure. This knowledge helps streamline the growing process to make it faster and more efficient.
The current way of tracking plant phenotypes is through manual images and annotations. Researchers have to study and observe the plants and note their characteristics themselves. This way of tracking data is much more time-consuming, and less efficient due to the innate human error of categorizing data.
The Benchbot will replace human efforts of imaging and categorizing plants by making the process fully automated. “The basic reason that we develop machines is to make human work easier, especially repetitive tasks,” said Priya Jakhar, an electrical and computer engineering Master’s student working on the Benchbot.
The robotic system is equipped with a high-resolution camera that takes multiple images of the crop. These images are sent to computer-based machine learning algorithms that are programmed to recognize and sort plant phenotypes.
The Benchbot provides an interdisciplinary research experience for students who are involved. Not only do they get exposure to the agricultural world, but electrical and computer engineering students can practice many different practical skills.
Students can work on the hardware side of the project which includes sensor integrations or incorporating embedded systems. On the software side, students can learn more about computer programming or machine learning algorithms. Additionally, the Benchbot project employs mechanical engineering students to help build the seven-foot frame and help with its automated mobilization.
“Ideally we want to be able to scale up with the benchbot… we want to be able to take it to a more realistic setting where we try to understand what are the factors that are impacting the responses in the different plants?” said Edgar Lobaton, an associate professor in the Department of Electrical and Computer Engineering.
There are a few Benchbots that have been built and deployed across the country to farmland. Here, the robot can practice real-world applications of monitoring plant phenotypes and weed species.
“For agriculture, right now, the primary revolution going on is the use of computer vision… we all see a future where robots are going to be doing more and more of the farming and computer vision systems are a key part of making that operate,” said Chris Reberg-Horton, professor of crop and soil science.