Smart Camera Tells Tobacco From Marijuana
An anonymous reader writes "A new smart camera technology not only takes a picture but also assays chemical composition, allowing photographers to tell whether that hand-rolled cigarette contains tobacco or marijuana. Designed to speed industrial inspection systems — such as detecting whether food is spoiled — the new smart camera includes spectral filters that make images of corn fields appear differently from hemp. Spectral cameras have been available for decades, but this microchip version should be cheap enough for almost any application."
From TFA:
The system-on-chip (SoC) solution can accurately distinguish between objects that appear virtually identical using traditional red-green-blue imaging chips.
The sentence immediately preceding that one, claims the product senses outside the visual spectrum ("hyper-spectral") and that it can perform remote spectral analysis, but somehow it uses just a good ol' RGB sensor.
Yes, it says that it can differentiate things that a traditional RGB sensor cannot. That means it's NOT a "good ol' RGB sensor".
Color cameras are just black and white ones with a set of filters over the pixels. Traditional color cameras use red, green and blue filters in a Bayer pattern. You can make a "hyperspectral" camera by using narrower filters of specific wavelengths to detect light at those wavelengths. For example, if you know that corn and someone else differ at a certain wavelength, use a filter at that wavelength.
You can also make a hyperspectral line imager by using a slit instead of a round aperture and putting a grating or prism behind it. That turns the slit image into a two-D "image" where the slit is broken down by color. One dimension is along the line, the second is by color. Move the camera so the slit covers the desired imaging area and record the spectrum at each "pixel" in the resulting image. Google for "CAP" and "Archer".