Over the past few years, we have seen significant growth in interest in our hyperspectral imaging technology. We hear about applicability of this technology in new areas almost every day. It may not be unrealistic to expect that in a few years we may all have mobile devices equipped with hyperspectral sensors used to monitor our health, what we eat, how we shop for our food, and more.
Hyperspectral imaging is a technique that captures and processes multiple narrow-band images over a spectral range, enabling detailed analysis and identification of objects based on their wavelength fingerprints. This technique -- sometimes referred to as multispectral imaging or even chemical color imaging -- essentially extends classical point-based spectroscopy with imaging, while extending classical grayscale/color imaging with more spectral data.
There are many applications for hyperspectral imaging. It started with use in remote sensing applications. More recently, we have spoken with specialists, from doctors to farmers to companies that sort fruits and vegetables, to UAV operators and even groups aiming to make a Star Trek tricorder-like device, all of whom are seeing the potential of hyperspectral imaging. Optical sorting can be used, for example, in mechanical sorting of apples, process monitoring, medical diagnostics, and even consumer applications.
We have heard from agriculture specialists who see value in this technology to monitor various parameters of soil and crop growth to control the conditions of fertilizer and water distribution. In another instance, a neurosurgeon reported this technology might enable him to perform tumor resections more precisely and avoid residual tumor regions. Almost as exciting as the technology itself is the experience of talking with all these people, as it is quite rare to see an enabling technology making such a broad impact on such diverse areas.
A common trend among all these uses is a clear need for inexpensive, compact systems able to process large datasets sufficiently at high speeds. Various innovations in spectral unit and camera designs are underway to meet this need. Imec is contributing to innovation in this area by monolithically integrating spectral filters with image sensors, enabling mass manufacturing of hyperspectral sensors and cameras while keeping the cost of production low and making the overall system very compact.
Interestingly, the dynamics in the supply chain are also changing. Traditionally, OEMs source image sensors, spectral units, and cameras from different suppliers to create custom data analysis products. However, as a result of monolithically integrating spectral filters with image sensors, we are seeing interest from image sensor vendors in moving higher up in the value chain by providing hyperspectral sensors, instead of generic image sensors without spectral filters.
Likewise, we are also seeing camera builders beginning to test the waters by offering hyperspectral cameras instead of generic cameras. This could be a significant shift for camera vendors, because they would not only need to provide software to read out the raw RGB/grayscale data, but also software to perform pre-processing of spectral data.
At the other end of the value chain, we see players in the pure spectroscopy market that traditionally focused on spectral data analysis of single-point-based measurements, including imaging aspects in their systems. This means they need to extend analysis from a single-point spectrum to simultaneous analysis of a 2D array of points, and spatial information becomes relevant.
Finally, making hyperspectral sensors that can perform meaningful analyses is not trivial. Thus we are also seeing the emergence of new companies that provide software and services to make sense of seemingly complex hyperspectral data.
This article was originally published on EBN's sister publication EE Times.