Imec’s prototype hyperspectral camera can capture multiple types of data about materials or different parts of the same object. This could help in the development of automatic object classification systems that rival those used in state-of-the-art hyperspectral references and recorded spectra of plant material. The innovation could lead to small, cost-efficient cameras that can be adapted easily into vision systems.
The prototype chip's hyperspectral filter, which Imec developed, has 100 spectral bands between 560nm and 1,000nm. The filter bandwidth ranges from 3nm at 560nm to 20nm at 1,000nm, and the transmission efficiency is approximately 85 percent. Under illumination by a 450W halogen light, the prototype's typical image integration times are between 2 and 10 milliseconds.
Imec has been working on the development of a hyperspectral sensor for some time, based on its research reports from 2010 and 2009. Some of those efforts targeted machine vision, medical, and security applications.
The speed of Imec's demonstration system corresponds to an equivalent line speed of 2,000 lines per second -- much faster than current hyperspectral sensors. To adapt the technology to different industrial vision application requirements, a different commercially available or custom image sensor could be substituted with different pixel sizes and frame rates. The number of spectral bands and the spectral resolution of the hyperspectral filters can also be changed.
What I like most about this technology is the huge difference in size between other multispectral cameras I've written about in the past and the fact that this is a chip-level solution, even doing post-processing filters on-chip. I think the need for this technology will only continue to increase as design features keep getting smaller, and with the mixes of multiple material types.
There's a large number of apps that could take advantage of this technology. Industrial machine vision and inspection of chips, boards and electronics sub-assemblies, R&D of several different kinds including component failure and analysis labs, medical labs of various kinds, and medical equipment manufacturing. It could possibly also be used in various kinds of materials detection, possibly in security apps, as well as for detecting counterfeit components made of inferior materials.
Hi Ann, this camera systems looks very interesting and powerful for many applications. I want to ask what is the typical spectral range of vision this components have and what is the difference with this one, I work with concentrated solar power systems and I wonder if this devices would work for analysis of ray tracing for beam radiation. Thank you and great article.
Aldo, I'm not exactly sure what your question is. As the article states, the prototype chip's hyperspectral filter has 100 spectral bands between 560nm and 1,000nm. The filter bandwidth ranges from 3nm at 560nm to 20nm at 1,000nm, and the transmission efficiency is approximately 85 percent.\
Interesting that it may be able to detect counterfeit components made of inferior materials. Right now, components coming back as returns are inspected by the human eye. There is a training program and certification for inspectors. But that process can't catch everything. Returns are the vulnerability area that lets counterfeit components into the legitimate-component bloodstream. A camera that can see better than the human eye could be a big help.
Having seen spectroscopy systems in the semiconductor industry in the 1980s, this seems like about as small a package as I can ever remember. Is this indeed smaller than the current state of the art? Has anyone else used a system on a chip approach like this one, Ann?
Chuck, there are other multispectral sensor chips of varying sizes, architectures and wavelength ranges--usually IR or IR plus visible light--but Imec's combination of hyperspectral spectroscopy sensor with regular visible light image sensor is unusual. It may also beunique. It's certainly one of the smallest I've seen, but I don't read German and I wouldn't be surprised if there are others (Germany is a leader in machine vision and imaging technology). That said, European Imec is a leading edge research institution, and they have a lot of firsts to their name.
Thanks, Ann! This is amazing technology. Taking a look at the original release, Imec mentions integration times as short at 2 ms, which, not counting transfer time, would equate to their specified max of 500 fps. When it comes to their hyperspectral filter, it has a maximum of 100 band settings at an effective slew rate of 2000 bands per second, which puts it around 500 us per band -- darn fast. I'm guessing that is the reaction time of the electro-optics they are using for their proprietary filter. Either way, this is a really impressive piece of technology that should find rapid acceptance in all sorts of machine vision applications.
William, thanks for the input. I'm not a specialist in this area, but I have a retentive reporter's memory and those specs sure looked leading edge, if not bleeding edge, to me. The one that first caught my eye, besides the integration of spectroscopy plus imaging on a chip, was the 6x maximum frame rate of the 4 MP image sensor. It's good to hear that the hyperspectral sensor frame rate is also wicked fast. The one thing we didn't get info on is the price. Since this is a prototype, it may go through some design shrinks for the first production runs that will help lower that.
It took Foveon 5 years to commercialize their technlogy of using a single CCD to get the RGB layes right. Unless this company is willing to publically publish a responsivity chart I am not weilling to hold my breathe. To me this is just another marketing department bringin infiormation way ahead of their enigneering department. I have seen it so many times in the past I have just given up on these press releases.
Show me the Specs or get out of my way. I got a job to do with real products in real time.
MicroImaging, thanks for your comments. I agree that there's a lot of vaporware out there. But this is not a breathless, hype-y startup. It's Imec, and they don't do vaporware. This is also a prototype, as we clearly stated, and as we all know, it can take time for a prototype's promise to become a reality, and not all prototypes become products in high volume on a production line. If you succeed in getting that chart from Imec, please let us know.
The second page, which has the spectral response information, was slow to appear, so some may have missed it, as I almost did. There may be a few non-machine vision applications, depending on the price of the camera. I do see some very interesting products if the price is right. But just as others have said, announcing the product is a lot different than shipping the product. So please be sure to make a big deal out of the announcement that they are shipping these cameras in any quantity.
William, thanks for your comments. As williamlweaver points out, this new R&D development has some incredible advancements. As to commercialization, that would come in the form of other companies incorporating the chip into their products.
At this year's MD&M West show, lots of material suppliers are talking about new formulations for wearables and things that stick to the skin, whether it's adhesives, wound dressings, skin patches and other drug delivery devices, or medical electronics.
Researchers at Lawrence Livermore National Laboratory have published two physics-based models for the selective laser melting (SLM) metals additive manufacturing process, so engineers can understand how it works at the powder and scales, and develop better parts with less trial and error.
Materials and assembly methods on exhibit at next week's MD&M West and other co-located shows will include some materials you should see, as well as several new and improved processes. Here's a sampling of what you can expect.
The Food & Drug Administration has approved a 3D-printed, titanium, cranial/craniofacial patient-specific plate implant for use in the US. The implant is 3D printed using Arcam's electron beam melting (EBM) process.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.