Machine vision is one of the strongest growing industries in German and European mechanical engineering. According to the latest figures from the VDMA in Germany, the companies in this industrial sector generated 18 percent more turnover in 2017 than in the previous year. And also all over Europe the increase was significantly above the average increase in turnover in mechanical engineering.
Hyperspectral Imaging represents a growth segment here. The potential of this relatively young discipline, Hyperspectral Imaging (HSI), is estimated to be massive for the future. Markus Burgstaller, CEO of Perception Park, describes the key difference to traditional vision systems as follows: "Hyperspectral systems offer a spectrum for each object pixel instead of a monochrome or colour value compared to traditional machine vision systems. Depending on the wavelength range and spectroscopic processing, high-precision colour coordinates, chemical material properties, but also coating thickness information can be derived from the spectral data. The output information of such a camera has a significantly higher degree of complexity, but also allows greater diversity and selectivity in terms of manageable applications." "Whereas traditional RGB cameras only depict the colours red, green and blue, hyperspectral cameras make it possible to distinguish between more than one hundred colours", confirms Dr. Jan Makowski, President of LuxFlux GmbH. "With such high-precision colour measurement the properties of materials can be examined and chemicals can be made visible."
Diverse fields of application
Tim Huylebrouck, Product Manager at Stemmer Imaging, indicates the type of possible applications behind this method: "Supposed identical objects can reflect completely different light spectrums owing to their chemical properties â€" stimulated by a broadband light source. They can then be distinguished with hyperspectral systems. No other machine vision solution can do this."
Dr. Georg Meissner, Managing Director of Specim, mentions one application as an example: "ZenRobotics is a leading global provider of robot-supported systems for waste separation. As the rubbish that is sorted in such plants often contains hazardous materials such as asbestos, it is important to safely and reliably identify such substances. That's why ZenRobotics relies on hyperspectral cameras from Specim, which have the necessary detection reliability, sensitivity and speed for this task."
Gion-Pitschen Gross, Product Manager at Allied Vision, also sees the area of recycling and sorting of plastics as an important application for hyperspectral systems: "HSI makes possible the automatic separation of plastic parts, for example polyethylene and polypropylene, which can be detected and separated on the basis of their chemical composition. In addition to existing colour sorting, materials can also be distinguished by their molecular properties. The quality of results in the sorting process is thus significantly increased." According to Gross, the inspection of foodstuffs also offers huge potential for the HSI technology: "Meat, fat and bones have different molecular properties which can be clearly recognised in a HSI image. This also applies to other materials which have hardly any differences in the real image, like the almost identical foodstuffs sugar, salt and citric juice." For cameras in the visible spectral range it is also difficult to detect physical changes to objects. This plays a big role, for example in the food industry, when fruit or vegetables are to be inspected for their degree of ripeness or possible mould. Here HSI systems offer suitable solutions, which Allied Vision addresses with its hyperspectral cameras in the Goldeye series, among others.
Daniel Hofmann, CEO of the Spanish company Photonfocus-Tochter Solpi, expects a breakthrough in the future of HSI applications with mobile carrier systems such as Precision Farming by means of UAVs (Unmanned Aerial Vehicles): "Camera systems can be installed on a drone, for example, in order to enable photogrammetry or inspection applications. Such a camera system may consist of several hyperspectral cameras, a GPS system, an embedded computer and lots more. The pictures taken have accurate GPS data to simplify subsequent machine vision." Solpi offers a camera system that allows the use of several hyperspectral cameras in an independent grabbing solution.
Challenges of HSI Imaging
Despite this and other promising application examples, Hyperspectral Imaging is at the moment still one of the exotic disciplines of machine vision. One reason for this is the number of challenges that still have to be overcome before the technology can be used comprehensively. The currently still relatively high price for hyperspectral technology is also one of the main entry barriers. On top of this is the fact that the entire technology is not easy to understand and often requires in-depth specialist knowledge in the area of spectroscopy.
Stemmer Imaging Product Manager Huylebrouck also mentions the topic of lights and illumination as another challenge: "Hyperspectral Imaging does not work with the LED lights and illumination often used in machine vision, but with halogen lamps which emit a wide wavelength spectrum. There is still a need for suitable illumination here." In addition, the lights must have protective glass, e.g. in applications in the food industry, which cannot be made from actual glass in this industry because of safety standards. "However, other materials distort the spectrums. A few tricks are needed here", explains Huylebrouck.
According to many experts, these reasons as well as the lack of high-performance hyperspectral software, reliable spectral data and experience are leading to a somewhat hesitant expansion of this new technology at the moment.
Trends and further developments
Nevertheless, the innovative opportunities of the technology are prompting many companies to work intensely on further developments in this area. "We are observing a trend towards the reduction of systems, whereby it must be ensured here that this is not at the expense of the performance. The future will indicate what limits are set here" states Hilmar Kruger, Sales Manager at inno-spec.
Perception Park CEO Markus Burgstaller lists some other current approaches: "In terms of other machine vision technologies, the trend is heading towards Embedded also for Hyperspectral Imaging. The cameras are becoming smaller and more affordable and in combination with new machine vision technologies will allow use in handheld devices like future smartphones in the foreseeable future." With the addition of a pre-processor HSI cameras are also becoming "smart" and make possible the pre-processing of the hyperspectral data volume as well as the extraction and transfer of the relevant information such as chemical or physical object information per object pixel. "This will lead to much greater acceptance thanks to the possible standard interfaces", states Burgstaller with confidence.
The combination with approaches from artificial intelligence and especially the topic of Deep Learning will also advance the technology significantly, believes Burgstaller: "In the future HSI systems will learn by means of chemical and physical information and thus ensure greater simplification of the application of Hyperspectral Imaging systems." Gion-Pitschen Gross from Allied Vision confirms this belief: "In the future it should be possible to detect materials solely based on their spectral signature, without the need for training." The hyperspectral software Perception Studio from Perception Park already provides an intuitive software suite for the recording, modelling and analysis of hyperspectral data.
For Specim President Dr. Georg Meissner, Hyperspectral Imaging is increasingly becoming a widely used and established segment of machine vision and quality inspections as a result of these numerous trends. "The technical advances in this area will very soon lead to higher image capture rates as well as probably also wider spectral ranges and more compact camera sizes."
For Florian Niethammer, VISION Team Leader at Messe Stuttgart, Hyperspectral Imaging is therefore one of the most exciting and fascinating topics of VISION 2018: "HSI systems are extremely innovative and open up new applications for machine vision which up to now could not be realised with traditional systems. The topic will therefore play a key role at this year's VISION fromÂ 6 to 8 November 2018Â in Stuttgart. In addition to numerous exhibitors, who will present their HSI products and solutions, several presentations within the framework of "Industrial VISION Days" also give interested visitors the opportunity to obtain information about this technology and generate ideas for possible applications."