A team of scientists from a Bosch startup called Deepfield Robotics recently developed an artificially intelligent and flexible agricultural robot that could revolutionize current methods in crop farming, as well as reduce the environmental impacts of weed control.
The compact car-sized robot, known as the BoniRob, uses laser positioning and satellite navigation to find its way around the fields, and it could detect its location to the closest centimeter, developers said.
Researchers estimated that farmers need to produce 3 percent more crops per year to keep up with population growth. In a paper presented [pdf] at the Intelligent Robots and Systems or IROS 2015 International Conference that occurred during Sept. 28 to Oct. 2 this year, developers explained how the BoniRob could contribute to novel innovations in technology.
Image Recognition Is Key
The BoniRob could render methods of plant breeding to be more effective and protect crops from weeds, researchers said. It could detect the shape of leaves and distinguish the difference between weeds and crops through image recognition.
Then, using a rod, the robot mechanically clears up weeds, and not with chemicals. This is why experts believe that the use of herbicide in farms could be ended as farmers begin to use the BoniRob.
Researchers explained that the BoniRob was designed after previous studies by plant scientists where the latter analyzed which crops could be improved. These scientists' tests were all done through manual lab work. In German, this plant assessment process is called Bonitur, and this is where the BoniRob gets its name.
With the BoniRob's automatic image recognition and an algorithm that analyzes photos, effort and time originally allocated for plant appraisal will be saved, scientists said.
"We are leveraging our expertise in sensor technology, algorithms, and image recognition to make a contribution to improving quality of life, even in areas that are new for Bosch," explained Deepfield Robotics General Manager Amos Albert.
In Case Of Confusion
However, there is a challenge. During the early stages of development of chamomile and carrots, their leaves look very similar, and this could become a problem for the BoniRob.
The solution? Albert and his colleagues made use of machine learning which involves a huge amount of image files where the weeds are highlighted.
Albert explained that the robot will learn to differentiate more and more precisely between plants that they want and the plants that they do not want based on properties such as leaf shape, size and color.
In the meantime, Deepfield scientists are still conducting real-world tests on farms with the BoniRob, but they intend to sell or rent models out to farmers.
Watch the video shown: