The cooperation achieved in the Phenomics Consortium would seem especially difficult to achieve. After all, what is the point of assembling together people from technology and defense corporations, agriculture companies and academic institutions? And what common language can high-tech professionals hope to find with farmers not to mention academics?
Yet, more than a year after launching Phenomics, no-one doubts the common vision any longer.
In order to understand the full picture, let's take a step back and learn a bit about this specific field – precision agriculture. Precision agriculture is an approach to agricultural management based on observation and precise scientific measurement of plants in agricultural growing conditions and, according to the results, providing a precise response of focused treatment.
Sharone Aloni, Senior Director of Sensing Systems at Elbit Systems, explains why the field of precision agriculture is presently at a critical point: "There is a global food crisis today that will only worsen with time. The field of improving ability to produce agricultural crops is therefore of great importance. Governments and broad-based organizations such as the UN devote significant attention to these areas.
"The State of Israel", he says, "can introduce many of its capabilities in this specific field as well as the immense knowledge that we possess in agriculture and sensors – things that Israel is renowned for worldwide. These are all reflected in the Phenomics consortium."
Today's global bottleneck in plant improvement and in research of food enhancement is in the area of phenotyping – the ability to perform a computerized identification and processing of data on the morphological and physiological traits of plants.
Dr. Hagai Karchi from Evogene, one of the company's founders and chairman of the Phenomics consortium explains: "The current ability to measure plant traits (phenotypes) is at the level of 19th century medicine. This is still a profession that is transmitted from one expert to the next as an art rather than as an exact science. Plant breeders for example, are learning to look at plants, understand their state and how to improve them.
"Furthermore, plant characterization is typically performed manually by human experts and not by machines in other areas of modern agriculture too. The main reason is that a high level of experience and expertise is needed in order to become familiar with the plants and their various states. Phytopathologists, for example, who are experts in plant diseases, need a deep understanding of the range of different diseases that harm each species in order to know how to identify the plants' relevant symptoms. Just like doctors, it's not enough merely to determine if the plant is sick or not, but rather, they need to be able to identify the disease itself, the stage it's at, and its intensity.
"Because this is a characterization that requires human expertise and which is difficult to perform on a large scale, an extremely large disparity has developed in global agriculture between what we would like to be able to do with regards to scale and precision, and what is possible from a capability perspective. This is the focal point of the Phenomics consortium – the ability to perform this characterization in a computerized manner with sensors, images etc.
"The idea to establish the consortium, with the assistance of the Innovation Authority, arose because of the global revolution in the field of computerized picture analysis and the parallel trend of increasingly smaller and cheaper sensors. We thought that if we choose the field of advanced sensors with computerized technologies for processing and analyzing pictures, we could precisely measure plants' morphological and physiological traits in their natural growing conditions, thereby in practice developing a virtual super-expert for plant phenotype characterization.
"Future technology such as this would have countless applications. One example will be the very early identification of diseases, even before they are visible to the eye, before the onset of symptoms. Such identification will enable us to rapidly treat the disease at its early stages in a manner that will negate the current need for extensive spraying of toxic substances. In other words, a healthier process for both the environment and humans as well as a financial saving for farmers. Another example will be the possibility to tell the growers exactly which fruit on which trees are ripe, thereby facilitating optimal harvest."
"Innovation blazes brightest precisely at the point where technologies from different fields come together. The MAGNET consortium programs attempt to connect different technologies in which Israel is leading the way," says Dr. Aviv Ze'evi, Director of the Technologies Infrastructure Division at the Innovation Authority. "At Phenomics we connected the sensor industry, developers of agricultural knowledge, and advanced capabilities in the field of data analysis in order to create an economic impact that can support all of the industries together or each of them separately."
A Common Language instead of the Tower of Babel
Dr. Zur Granevitze, one of Phenomics' founders and the consortium's current technological director, tells how it was founded: "There is currently no solution for phenotyping technology in agricultural growing conditions and we identified a tremendous commercial-technological need. Even before the consortium was established, we focused on clarifying the needs and technologies of the various companies. When you gather companies, each of which has its own interest, its essential to create a win-win structure during the development process. This is also the foundation for cooperation between the teams from industry and academia – and within the academic and industrial teams themselves. The Innovation Authority deserves much of the credit for assisting us in this clarification process.
"We have been fantastically successful in creating this model at the Phenomics consortium. This success is because there is no clash in business interests of the companies from the agriculture and defense fields and because each agricultural company operates in a different segment so there is also no competition over technologies. This way, each company benefits from the consortium without creating a conflict with the parallel companies.
"The managerial structure that we designed at the consortium, is based on seven separate "working packages" each of which develops a critical component for the system and deals with gathering the data, its analysis and management. Each team contains people from different companies and academic groups and a director who coordinates the activity. The connection between these "packages" is maintained via the directors who advance things together as a community on a technological and managerial level.
"Dr. Karchi adds: "All seven teams work together toward a common goal: development of technology that will enable us quality identification of plant performance, from which we can then develop applications that are important to farmers worldwide.
"One of the challenges was to teach the people at Elbit, Opgal, and the others about plants and then teach the deep learning data analysts and researchers which traits we want to measure and what's more and less important. The process of creating a language for people from very different fields was a very big challenge. Instead of the Tower of Babel, we created a common language between physicists and engineers, computer people from academia, and plant experts.
"It's difficult to explain to someone else about the enormous difference that exists between people from the industry and academics. Those of us in the industry want to create an infrastructure for a product or the product itself while in academia, they want to create an academic publication. The issue of language and culture frequently divides people from the applicative field who possess a practical approach and are working within a time constraint, and people for whom time is not the most important criterion and whose objective is to create knowledge.
Dr. Daniel Koster, a team leader of Core Technologies at Hazera and Coordinator of Data Analysis at the consortium, also relates to the special human composition at Phenomics: "This is a very interesting subject, partly because our fields are so diverse – professors in computer science, agronomists, physicists from the optics field and professionals from the defense industry. This is undoubtedly a very heterogeneous consortium. It is challenging, but with everyone's good will, it works well. Although everyone at the consortium comes from different companies and a different culture, there is already an atmosphere of togetherness, and this plays a large role in our success."
When the Camera Moves from the Tank to the Field
Dr. Ilya Leizerson, Director of Innovation in the C4I and Cyber Division at Elbit Systems, explains the company's involvement in the consortium which seems far removed from the fields in which it normally operates: "The subject of precision agriculture is important to Elbit from several aspects. First, in order to expand the company's range of capabilities into the civilian sphere. The consortium's activity connects the capabilities developed at the C4I and Cyber Division at Elbit Systems to the world of agriculture. We are developing simple and easy-to-use systems that are required to operate in complex field conditions and to provide a year-round response. We are utilizing the innovative technological capabilities that we have developed over the years to identify plant traits with tools that were not even thought of previously. The new technologies are deployed in sensory and remote identification tools, different advanced optical methods and extremely advanced algorithmic capabilities.
"We developed a calibrated system for the consortium that measures a range of plant traits, in which all the sensors look at the same series. The first system we developed was operated manually and currently we are already working on the third generation with the hope of developing a system that operates on an autonomous platform – initially on land and then in the air. The ultimate goal is technology for large-scale coverage without sacrificing resolution and precision.
"Our goal is for the system to replace slow manual processes that require highly skilled experts and to use an automatic, computerized, and controlled process that will assist farmers to make better, more precise, and cost-efficient decisions regarding plant treatment."
Dr. Karchi adds: "The cameras have eyes just like people, but they see additional spectrum fields invisible to the human eye. It turns out that the area of "bionic eyes" is crucial for agriculture because symptoms of disease and stress for example, initially appear in spectrums that are undetectable to the human eye."
The next stage is that deep learning machines will teach themselves how to read and analyze pictures of plants. For that to happen, we need a very large database of plant parts pictures as they look in the field and not in the lab. "We currently have teams of photographers from the three agriculture companies in the orchards, the fields and the greenhouses. All the pictures are being collected in a single database at Ben-Gurion University in a precise catalogue format that can be accessed by all the participants in the consortium", says Karchi.
Dr. Koster adds: "Only the agro experts possess the expertise necessary to judge the plants' traits and they bring these capabilities to the consortium. We receive pictures from the field with expert analyses and that's where the subject of deep learning picture deciphering comes in play. This is a relatively new field that has taken off significantly over the last five years and which is influencing many other fields: autonomous vehicles, healthcare and others.
"We adopted deep learning for agriculture: we show the computer a huge number of plant pictures together with the interpretation provided by human experts, so that in the next stage, the computer will be able to decipher them by itself.
"A large number of different parties take part in the analysis. The analysis team includes researchers from Ben-Gurion University, Haifa University, the Technion, and from the Hebrew University. Together with a development team from the ag-tech companies, we search for new ways to contend with the challenges involved in deciphering plant traits received from multi-channel data: RGB channel, depth channel, multi-spectral channels, and a thermal channel.
"The objective is to extract quantitative values of the plants' phenotypes from the data, including the presence of a virus, the fruit's size or degree of ripeness, the presence of abiotic stress and more. A large part of the challenge is that unlike pictures taken in a controlled lab environment, pictures taken in the field are difficult to decipher because of the changeable conditions throughout the day, winds blurring the image of the plant, parts of the plant that are hidden from view or remain in the shadow, and plant parts that are positioned in non-ideal orientations for imaging. In reality, the plants simply don't line up nicely in front of the camera like in the lab and this poses difficulties for the algorithmics", Dr. Koster summarizes.
A Dynamic Spiral Process
Dr. Koster describes the work together in the consortium: "Hazera's objective as a member of the consortium is to both contribute to it and benefit from it. To work together with the other members in one consortium is a tremendous privilege and not a trivial achievement. We benefit greatly from the participants' experience, the collaborations with them, and from the in-depth understanding of this field in academia. We have already completed the first year and so far, everything is working excellently."
Dr. Orna Livneh, Global Innovation and IP Manager at Hazera and member of the Phenomics consortium management: "The consortium offers things that can be leveraged as part of new initiatives beyond the regular activity of each company – the opportunity to establish startups for example. Furthermore, we utilize the consortium's products for our own use in developing our own products – which will be faster, more precise and better targeted."
Dr. Leizerson adds: "The consortium is supposed to exist for three years however the future commercial potential is also important. We are focused on results, so we work using agile processes – every development is immediately scrutinized in the agricultural sphere in order to proceed more rapidly with optimal adaptation to agro companies. The manual systems we have developed already enable agronomists to gather data. The next stage is a platform on a tool with sensors that will gather data with a higher dataflow rate."
Dr. Granevitze summarizes: "In this consortium, the inter-dependence between the partners is extremely high with immediate feedback. The dynamic spiral process created a very close and special bond between the consortium members. In the absence of commercial conflict, the cooperation is extremely impressive. We have established a transparent computational platform that was developed for all the consortium members and everyone works well together. I have even heard from highly experienced people that the degree of cooperation between the consortium's companies is much better than what they see within their own companies!
The development process at Phenomics is no less than amazing – after just over a year, we already have advanced and impressive sensory systems that are manufactured by the defense industry integrated with dedicated set of algorithms and supported by the consortium's computational platform. Agricultural thinking is in the heart of these technologies which are implemented by agriculture members of the consortium. External companies from Israel and abroad have already expressed great interest in the consortium and joint activity is currently being evaluated. We have also been approached by embassies that represent countries who have heard of our activity and wish to learn more about the initiative."