Hyperspectral Imaging to Change Agricultural Production

Article By : Anne-Françoise Pelé

A hyperspectral remote sensing center is being launched to power precision agriculture, mining and infrastructure inspection applications.

From when to plant to when to harvest, farmers have a plethora of decisions to make. The agricultural sector is undergoing a data-driven transformation, and farmers are adopting data analytics and smart farming technologies to improve profitability and risk management. Remote sensing, which generates field-observation data from satellites, manned or unmanned aircrafts and terrestrial sensors, is now a backbone technology for precision agriculture.

The global agriculture analytics market accounted for $590.03 million in 2018 and is projected to reach $2.46 billion by 2027, growing at a CAGR of 17.2%, according to ResearchAndMarkets. An opportunity that Headwall Photonics Inc. (Bolton, Massachusetts) and Geo-Konzept GmbH (Adelschlag, Germany) have seized with the creation of the Center for Hyperspectral Remote Sensing Europe (CHRSE) for the development of airborne hyperspectral and lidar technology and applications.

At the plant level

There is no doubt that hyperspectral remote sensing can change agricultural production into a more sustainable business. Because it sees beyond the natural limitations of human sight, it can detect diseases early in the crop growth cycle and enable the farmer to take prompt and targeted actions.

But why is its adoption so slow? The main reason is that the technology is too expensive, and farmers don’t have a clear outlook on the return on investment.

“Hyperspectral imaging is not yet ready to be applied by a standard farmer,” said Johannes Kutschera, head of Mining, Remote Sensing Department for Geo-Konzept. Going back in history, “one GPS system cost about one hundred thousand euros (about $110,600),” he recalled. “Now when we talk about steering systems for tractors, you can get GPS systems starting at €5,000 (about $5,530).” Of course, it is taking some time, “but our long-term goal is to convert this technology into a tool that can be used by a standard farmer.” In the meantime, he said, “it will be our task to find applications where this higher value sensor can result in higher value for the customer.”

Another threshold to adoption, continued Christian Felsheim, European Area business director for Headwall, is the complexity of the technology. “That’s exactly what we, Geo-Konzept and Headwall, are dedicated to work on, […] to make the technology easier to use, and that will eventually bring the cost down. But that requires that you bring technology and applications together, and have a place where it is easy to try things and walk through applications.”

The CHRSE will support the implementation and utilization of hyperspectral imaging technology combined with other sensor technology such as lidar and high-precision GPS. “Headwall is a high-end technology company, and Geo-Konzept will be the bridge in transforming the high-end technology and bringing it to a level where it can be used by many customers in the fields,” Kutschera explained.

Geo-Konzept’s headquarters in Adelschlag, Germany

The new center will be located at Geo-Konzept’s headquarters in Adelschlag. The facility features large areas for unmanned drone flights and certified unmanned aerial vehicle (UAV) pilots available to test and demonstrate hyperspectral imaging technology in application-specific environments and to train the next generation of UAV operators. In addition to all of that, it’s a farm. The company’s CEO, Thomas Muhr, is a farmer. And a lot of its customers are seed breeders with allocated field trial plots around the office.

“Individual plots are just one by one meter,” Kutschera specified. “Collecting information about every single plot is a very time consuming task with a lot of manual work, and it is not very precise.” It merely notifies if the crop is “in good shape or not, whereas hyperspectral imaging gives us real measurement data,” he said. “This measurement data can be further analyzed in the software we provide to plan the field trials in advance, the same software we will use to analyze the data on a statistical basis. We can collect information at a higher rate, maybe every week or every day when necessary, and with higher precision and good repeatability of the measurement.”

A plant dynamically develops phenotypes from its interaction with the environment. Understanding the plant’s growing process in an ever-changing environment is essential for efficient crop management and agronomic decision making.

“In plant phenotyping, there is a growing season, so we collect data over the full growing season, from March to July,” commented Kutschera. “At regular intervals, or when agronomists tell us it is a critical point in plant growth, we use hyperspectral imaging and apply different algorithms, neural networks, artificial intelligence, whatever is needed to find the mathematical solution for the problem.”

With hyperspectral imaging, a plant disease or a biotic stress can be detected at an early stage to prevent pre- or post-harvest losses. “There are signals that you can read from the plant, and that’s a good indication of a plant being potentially affected by a virus, another kind of pathology or a lack of water,” Felsheim said.

Using UAVs and hyperspectral imaging is an upscaling, as it provides detailed spectral signatures. “We are working on a different scale,” Kutschera said. With an aircraft or a helicopter, each pixel on the image is several feet or several meters in dimension, whereas “with a UAV, we can go down to the centimeter level and distinguish single plants, like a corn plant or a potato plant as it is growing. We can work on the plant level.”

This different level of detail, in combination with the 3D geometry of the lidar, provides many new opportunities on how to analyze the data. “A single sensor is not enough to fulfill the needs of the market,” Kutschera insisted. “In the past, we already supplied thermal images and multispectral images to our customers, but the hyperspectral imaging delivers a lot more information. And combining hyperspectral imaging with lidar, thermal imaging, even with soil data and chemical analysis data of the plant, etc. will bring us one step further.”

Other fields

Agriculture is the primary focus, but not the only one. The center will also work on mining and infrastructure projects. Vegetation management is indeed a complex and costly part of utility and infrastructure operations. Bushes, invasive species, unhealthy or overgrown trees can imperil power lines and towers, but also damage or block roads, railways and bridges after an episode of rough weather.

Detection of vegetation encroachment on railways (Image: Headwall Photonics and Geo-Konzept)

As part of an application-specific mission for the Deutsche Bahn Rail network and with the permission of DB Fahrwegdienste GmbH, Headwall and Geo-Konzept could fly a drone over railways and take hyperspectral and lidar images. The resulting 3D images provided useful information on plant health, height measurements of the vegetation and the position of potentially at-risk trees adjacent to the railway. This eliminated the need to send arborists, who would normally walk or drive along the railways. The drone could fly at will and identify areas at risk, saving some time and money, partners said.

It is about “having the opportunity, with little effort, to remotely be able to scan a large area,” said Felsheim. Unmanned aircrafts can also cover areas that were formerly difficult or impossible to go to and fly at lower altitude for higher resolution and better quality images.

“Flying a drone is not allowed in every area, and you need the equipment,” he summarized. “Having it all in one place, bringing people together to create new applications and find new ways of utilizing this technology is really the idea behind the remote sensing center.”

Subscribe to Newsletter

Test Qr code text s ss

Leave a comment