Oct 01, 2022
Share this article

Israel was celebrated during the first decades of its statehood for being a modern agricultural power. Since then, as apartment buildings replaced much agricultural land to accommodate the growing population and the economy had to become more sophisticated, hi-tech industries led to the country being transitioned over the years into the Startup Nation. It turned out that the connection between the two – hi-tech and agriculture – Agritech – is likely to give the world a significant boost in food production. 


This is the reason for the establishment of the Phenomics Consortium, sponsored by the Israel Innovation Authority. The consortium was created with the goal of furthering scientific and technological innovation through collaboration between academic research institutes and industrial companies.


The name “phenomics” is derived from the word “phenotype” – the observable physical properties of an organism, which possess agricultural, agronomic or biological significance. In this case, it refers to the diagnosis of the plant’s condition on the basis of its observable characteristics – color, shape and size. High throughput automated phenotyping today accounts for the bottleneck in the improvement of agricultural crops, so this shows the importance of the new study.

Researchers at the Technion-Israel Institute of Technology in Haifa have developed an innovative technology for automated monitoring of stress in agricultural crops. Prediction of stress conditions is important for monitoring plant growth stages, disease detection and assessment of crop yields. Early detection of water and heat stress is crucial for farmers, since reduction in moisture is reflected in limited stomatal conductance (a measure of the degree of the opening of stomata – tiny openings or pores in plant tissue that allow for gas exchange – that can be used as an indicator of the water status of plants). The development was led by the people of [he Geometric Image Processing (GIP) lab in the Technion’s Faculty of Computer Science. 

The team – research assistant Alon Zvirin, GIP lab head Prof. Ron Kimmel and chief engineer Yaron Honen – have developed smart technology for the monitoring and prediction of crop stress and leaf segmentation. “The detection of drought stress enables the plant to be saved, allows for the identification of diseases and the prediction of crop yield quantities, all of which are crucial information for the grower,” Zvirin explained. 

Artificial images of tobacco plants (right) and two Arabidopsis species (middle, left). The top row presents the artificial images, and the bottom row – the leaf masks. By creating a large quantity of artificial (synthetic) leaf images deep neural networks can be trained, thus providing for better leaf separation in real photographs.

“Through the use of color photographs, thermal imaging and deep learning, the researchers were able to predict stress and new leaf development with great success; in a test of the technology on banana seedlings, an impressive prediction level of over 90% accuracy was achieved. In the context of the latter – leaf segmentation – the researchers achieved unprecedented results in the identification of Arabidopsis [small flowering plants related to cabbage and mustard] and tobacco leaves by applying deep learning. To train the system on a large quantity of samples, the research team developed a vast database containing artificial leaf images, and then also tested the technology on other crops – avocado, bananas, cucumbers and maize,” he continued. 


“We included young researchers who were just starting out in the technology development process. They brought excellent ideas and did a great job,” said Zvirin. The article on stress detection was published at the European Conference on Computer Vision, ECCV, and the paper on segmentation was published at the Conference on Computer Vision and Pattern Recognition.