With the price of hyperspectral cameras falling fast, the new imagery on the block is poised to have a big impact on food and agriculture among other industries. Hyperspectral imaging is actually 30-year-old technology, created by NASA but hyperspectral cameras have until recent years, been too large and expensive for most companies to consider.
Universities and the private sector are currently working to make the cameras smaller and cheaper. VTT Technical Research Centre of Finland, which produces one of the first consumer hyperspectral cameras, has also modified an iPhone to take hyperspectral images.
“Hyperspectral imaging technology is now being used in industry and research to determine temperatures, moisture, chemical compositions, sugar content and fat in the food industry as well for a range of applications in biology, defense, forensics, pharmaceuticals and industrial inspection,” reads a Zion Market Research report that values the market for hyperspectral Imaging at $11.34 billion across all industries.
Hyperspectral cameras work by emitting a certain spectrum of light, from the spectrum humans can see to near -infrared, and then creating an image of the light that reflects back. The reflection allows the chemical composition to show in the image. With the right baseline data and often the help of some artificial intelligence, these images allow algorithms to detect things no other image and certainly no human eye can.
There are two main areas where hyperspectral imagery is currently being put to use in food and agriculture, but more are likely on the way as this technology democratizes.
Gamaya is a drone sensor and analytics platform that uses a combination of space-borne and proprietary drone-based hyper-spectral imaging data, along with the corresponding historical climate and weather records to provide farmers with alerts about pests and disease, yield predictions, and prescriptions for input application rates. Its current platform services corn, soybean, and sugarcane growers.
Gamaya has developed its own drone-mounted hyperspectral sensor. The sensor is Gamaya’s enabling technology, however, and it says its main intellectual property is in its analysis of this hyperspectral imagery using artificial intelligence to produce information about the plant’s physiology.
FluroSat uses various remote sensing methods, including satellites, drones, and some aerial imagery, to capture and analyze hyperspectral images of cotton and grain fields to predict disease and help growers make decisions related to crop health.
Multispectral cameras can measure generic characteristics such as if a plant is healthy or not, but hyperspectral images can go one step further, and diagnose the exact reason for that state, according to FluroSat. That’s because the extra bands of light they can detect can be associated with specific physiological traits within the plant.
The FluroSat platform shows nitrogen maps and suggests exact locations for agronomists to take tissue samples. They then enter the results into the platform to further calibrate fertilizer recommendations. The fertilizer prescription maps that result can be integrated with other farm management software platforms in order to incorporate historical data. The company claims the platform can also predict disease five to seven days before the human eye can see it.
Bayer + Planetary Resources
In 2016, multinational agribusiness Bayer’s Digital Farming department began a partnership with satellite startup Planetary Labs. While the exact details of the partnership were not disclosed, it is understood Bayer combines the insights generated by Planetary Resources data with physical data collection products and in-house agronomic modeling to provide “specific, granular and robust field-specific recommendations to farmers,” Tobias Menne, head of Digital Farming at Bayer, told AgFunderNews in 2016.
Planetary Resources’ Ceres constellation satellite was one of the first commercial hyperspectral sensors in space and its next spacecraft includes the first-ever commercial a midwave-infrared sensor offering thermographic mapping and night-imaging, useful for soil moisture and crop scouting applications, according to the company.
Hyperspectral sensors can also detect the type of crop or weed growing as objects leave unique fingerprints or “spectral signatures” in the electromagnetic spectrum. This is also used to detect other materials such as oil.
While Planetary Resources focused on soil moisture and temperature readings in the short term, Lewicki told AgFunderNews the potential to measure more advanced issues such as pest stress would be further down the pipeline.
In 2017, Planetary Resources announced that it would set aside its earth observation initiatives in favor of its core mission: asteroid mining. Stay tuned as we learn more.
ImpactVision uses hyperspectral imaging to help food supply chain companies determine the quality and ripeness of food products. The startup uses third-party hyperspectral sensors to essentially photograph food and pick up on certain characteristics that indicate what condition it is in. In meat, ImpactVision is able to determine tenderness, enabling meat producers to guarantee the quality of their meat for premium pricing. Another example is avocado ripeness — avocados are typically sold at least a few days from optimum ripeness — which ImpactVision can measure through the imagery and thereby help retailers to sell avocados at a more optimum time for consumers.
ImpactVision trains its software using computer vision and machine learning to recognize what these characteristics look like in a hyperspectral image through ground truthing by comparing images to manual tests on food. It undertakes this ground truthing on the behalf of a client or in partnership with them whereby they contribute by uploading images and measurements such as pH data to a server.
Some characteristics, such as freshness in fish, can be determined from the image alone: fresh filets reflect more light, Abi Ramanan, CEO of ImpactVision, told AgFunderNews.
Currently, ImpactVision is using one main hardware provider for its work, but the plan is to be compatible with a range of hyperspectral sensors, which Ramanan sees as becoming increasingly commoditized.
Are you using hyperspectral imaging in you agrifood tech business? Let us know! firstname.lastname@example.org