Artificial intelligence and planting

Technology developed by USP researchers uses computer vision to evaluate images of leaves, allowing for nutrient deficiencies in corn crops to be detected mathematically

6/14/2011 By Fábio de Castro

Agência FAPESP – An experienced farmer can tell, by looking at the leaves of a corn plant, if it is lacking in certain nutrients. But this is only possible when the plant is already full-grown and the crop has already been affected.

New technology developed by an interdisciplinary group of researchers at the Universidade de São Paulo (USP) can make this diagnosis earlier, allowing the farmer to intervene in time to save crops and avoid losses.

The technology utilizes digital imagery of the leaves and computer vision methods. In just a few minutes it can detect a lack of many different nutrients in seedlings in the first stages of development.

The Visão computacional aplicada à nutrição vegetal [Computer Vision Applied to Vegetal Nutrition ] project, developed by researchers at the Physics Institute of São Carlos (IFSC) and the USP School of Zootechny and Food Engineering (FZEA) in Pirassununga, has already been patented and is in the technology transfer phase. Laboratory tests were successfully concluded and the team is now performing field studies.

Aside from professors Odemir Martinez Bruno from IFSC, Pedro Henrique de Cerqueira Luz and Valdo Rodrigues Herling from FZEA, other researchers on the project were post-graduate students Liliane Maria Romualdo, Fernanda de Fátima da Silva, Mario Antonio Marin – all from FZEA and Alvaro Gómez Zúñiga from the Instituto de Ciências Matemáticas e Computação (ICMC), at USP São Carlos.

Bruno and Luz received FAPESP funding through the Regular Research Awards Program, with the Computer Vision Methods Applied to Plant Identification Project and the Analysis and Evaluation of the nutritional status of corn plants using an artificial vision system Project, respectively. Romualdo, Silva and Zúñiga are FAPESP doctoral fellows.

According to Bruno, artificial intelligence specifically for recognition of visual patterns on seedling leaves was used. The patterns correspond to a lack of nutrients such as nitrogen, phosphorus, magnesium, sulfur and potassium, and of micronutrients like copper, iron, zinc, and manganese.

“The leaves of adult plants show visual patterns that correspond to the lack of each nutrient. In plants in the early stages of growth, one or two weeks old, the patterns are already there, but not visible. Our challenge was to identify them mathematically,” Bruno told Agência FAPESP.

According to him, the method is based on reading digitalized images of leaves with a scanner. Once they are read, the image is transformed into a mathematical model that is compared to previously established models by software.

“We built a mathematical model of the seedling leaves, with the ideal quantities of all the nutrients. Based on that information, the software produces a new mathematical model that can be compared to the ideal, identifying the deficiencies,” he explained.

Most farmers study the nutritional makeup of the soil before preparing it for planting. But this preparation, according to Bruno, doesn’t necessarily guarantee that the plants will absorb the nutrients present.

“Many times, even if the soil has been prepared well, the plant’s phenotype doesn’t allow for absorption of the nutrient. When the plant reaches adulthood, an agronomic engineer can tell the difference. But at that point, it can only be fixed in the following crop,” he explained.

A severe lack of nutrients can harm up to 50% of a corn crop. “The technology allows us to evaluate the plant at one or two weeks and therefore the producer has a number of months to recuperate it before production. With early detection of the problem, substances can be applied to the plants that allow them to absorb the necessary nutrients,” affirmed the researcher.

According to Bruno, tests performed in the Pirassununga laboratories showed that the system worked with 87% accuracy. The group is now repeating the same studies in the field in the same town (inland São Paulo State) on areas with problem soil.

“From an academic perspective, a great deal of testing still lies ahead. The project aims to advance science in this area and we have to study a number of other aspects related to corn, to a variety of nutrients and to the application to other cultivars,” he said.

However, from the technological research standpoint, the method is already in advanced stages according to the IFSC-USP professor.

“The technology is nearly ready for use. In practical terms, we still don’t have a tool to measure the nutrients in plants this way. Even if we have a long way to go scientifically, the tool is already ready to be used for many nutrients,” he affirmed.