Spectral imaging includes multispectral imaging (in 4 to 30 spectral bands) and hyperspectral imaging (in more than 30 spectral bands). In the context of Industry 4.0 deployment and worldwide concerns about food safety and sustainable production, the need for low cost, compact sensors, able to provide advanced measurement is high. The global spectral imaging market is expected to grow at a 20% CAGR in the next 5 years, reaching more than 9 000 cameras sold in 2022, according to a recent report from Tematys, “Spectral Imaging: End-user needs, Markets and Trends”.
Spectral imaging, with its ability to combine chemical and high resolution spatial information, meets this trend. From bulky and expensive systems used for satellite based earth observation, hyperspectral and multispectral cameras are now becoming portable and affordable sensors. The improvements on size and cost, added to the development of robust and application-oriented data processing solutions, are leading factors of the adoption of multi/hyperspectral for on-field and in-process applications such as machine sorting, plant health monitoring, food quality control, environment monitoring, etc. The highest adoption rates will be seen in the following markets: sorting in the recycling and food industries, precision farming and color machine vision.
Sorting in the food and recycling industries
Sorting was the first application of hyperspectral cameras in industry. In terms of market, it is currently the main application of spectral imaging. It accounts for more than 25% of the multi/hyperspectral cameras volume in 2017.
It has first been adopted in the 2000’s on sorting machines in the recycling industry to increase sorting speed and quality, compared to manual sorting. Hyperspectral linescan cameras in the NIRSWIR range (900 – 2500 nm) are used to image wastes at high throughput on conveyor belts. They are especially suitable for sorting plastics, to separate different types of plastics of the same color according to their composition. It is also used to distinguish printed paper - that needs de-inking - from non-printed paper like cardboard. Compared to classical methods (mechanical), spectral imaging provides a more advanced analysis of materials. It meets the need for high purity after sorting, to produce high quality end-products after recycling and increase profitability of recycled objects producers.
Spectral imaging sorting is now spreading to the food industry, at a rate of 24%.
Sorting of plastics with NIR spectral imaging. Courtesy of Specim.
Two factors ave been the main drivers of the growth of the spectral camera market: The decrease in the price of cameras to around $10,000-15,000, and the several food frauds over the last years (melamine in infant’s milk in China in 2008, horsemeat scandal in Ireland in 2013). Spectral cameras address the growing need for sorting of inputs. Indeed, in agrofood, the quality of input materials varies as the end-users, (i.e., the consumer), want a steady quality and taste. Hence, transformation processes need to be permanently adjusted. Hyperspectral NIR imaging is able to provide detailed information on the molecular content of fruits, vegetables or cereals (protein, sugar, water, etc.). Spectral imaging cameras are also able to detect contaminants that are complex to identify by visual inspection, like maggots in rice.
With the increasing number of regulations on waste management and on food safety, sorting in the recycling and food industries will remain one of the biggest markets of spectral imaging in the next 5 years.
Spectral imaging is used to assess crop conditions and optimize agricultural activities. A better control of the production and the automation of some tasks helps improving farms profitability while reducing environmental impact.
Spectral imaging is mainly implemented on remote sensing platforms to provide field chemical mapping. The first were satellites in the 1970’s, and, few years later, aircraft systems. It provides useful information on crop maturity and stress with a wide spatial coverage. However, the cost of spatial imagery and the flight-related logistics is high, and the spatial resolution above 5 m is too low for certain applications where crop rows are separated by less than 5 m.
Recently, the development of unmanned aircraft vehicles (UAVs) has given access to an affordable platform with enhanced spatial resolutions and higher operational flexibility. The current growth of spectral imaging in precision agriculture is a direct consequence of the booming of the UAV market. Drones manufacturers are now targeting professional applications like farming. But to reach this market, they first need to decrease costs of drones and to simplify their use (automated flight) to meet the demand of farmers for imaging on large areas, in short-time periods. The goal is that, ultimately, all farmers have an UAV platform just like they have their tractor. Second, they need to functionalize the UAVs with tools providing relevant information to end-users. Multispectral and hyperspectral imaging has a key role to play because it brings a high added-value for precision farming applications : it is able to assess several critical parameters like water stress, chlorophyll amount in leaf and plants, pathogens, etc. The availability of this information on a 2D map of the field allow the smart application of fertilizers and nutrients, optimized harvesting and irrigation times and early diseases detection.
The high demand for such valuable systems combining high spectral and spatial resolution has driven the decrease of spectral imaging cameras’ cost. Currently, mainly multispectral cameras (4 or 5 bands in the visible-NIR range where leaf chlorophyll content and vegetation stress can be detected) are used. They have reached costs up to $4 000-5 000, like Parrot’s Sequoia camera or the RedEdge camera from Micasense. The market of spectral imaging cameras in precision farming is expected to experience the highest annual growth rate – 34% in the next 5 years.
Color machine vision
Machine vision applications demand more advanced measurements like 3D imaging, high resolution or spectral imaging to support the development of Industry 4.0. Major machine vision camera manufacturers like JAI or Teledyne now offer multispectral products (4 bands: R, G, B and NIR). A few years ago, to meet demands from customers, Stemmer Imaging, a major machine vision equipment provider, built a line of multi/hyperspectral systems for OEM integration. The demand is expected to continue growing in the coming years, and lead to a market growth of spectral imaging cameras in machine vision of around 22%.
One of the major application segment of spectral imaging in machine vision is color inspection. The printing industry is among the main industries benefiting from multispectral cameras to monitor color variations on commercial publications, newspapers, currencies, pharmaceutical and food packaging, etc. More and more linescan cameras are implemented on processes manufacturing materials in roll or in sheet (textiles, thin film, plastics, coatings, …) to inspect color homogeneity. The automotive industry uses spectral imaging for the quality control of paint. Spectral imaging has also been adopted in the display manufacturing facilities. Unlike current off-line control techniques, they continuously test the conformity of the optical characteristics of the devices all along the production line.
Solving industrial and societal challenges
Several challenges remain for the wide adoption of spectral imaging on-field or in industrial line. In particular, the cost of cameras must decrease below $5 000 to meet the investment capacities of field professionals and industrials. However, the booming of this technology is getting closer, driven by end-user demands to solve major industrial and societal challenges.
Written by Clémentine Bouyé, Benoît d’Humières from Tematys.