As you might expect from its name, a hyperspectral camera does more than reproduce the appearance of objects. The name connotes spectroscopy: For each of its pixels, a hyperspectral camera yields a continuous spectrum over the same, wide waveband.
Given that molecules’ distinctive fingerprints lie in the IR, hyperspectral imaging is especially useful when the camera’s waveband encompasses both the IR and the visible. Indeed, IR-to-visible hyperspectral imagers have found a host of applications, including remote sensing of vegetation, prospecting for oil, and surveillance of criminal suspects.
Most hyperspectral cameras owe their spectroscopic ability to a diffraction grating, which spreads the light from a narrow slit-shaped aperture over a sensor. If the slit is oriented in, say, the x direction, then sweeping the aperture over a scene by means of a movable mirror builds the image in the y direction.
The narrow slit and long focal length yield fine spectral and spatial resolution, but at the expense of throughput (because the aperture is small), camera size (because of multiple optical components), and mechanical complexity (because the optics move). In 2009 Andy Lambrechts and his colleagues at IMEC in Leuven, Belgium, set out to design a cheaper, more compact camera. Last week at SPIE Photonics West, the camera was publicly unveiled for the first time.
Chips and filters
Besides its standard optics, the IMEC camera (shown here) consists of two main components, a CMOS sensor and a Fabry–Pérot or dichroic filter. The camera’s CMOS sensor is the commercially available 2048 × 2048-pixel CMV4000 made by CMOSIS of Antwerp, Belgium. It operates in the same way as other CMOS sensors: Each of its silicon-based pixels converts photon energy into electronic charge, which is then read out by the pixel’s transistor-based electronics.
The Fabry–Pérot filter was developed by Lambrechts’s team and is what makes the IMEC hyperspectral camera special. In cross section the filter resembles a staircase of tiny steps. Interference between the top surface of each step and the bottom surface of the staircase ensures that the step transmits only one spectral band to the sensor below.
The camera unveiled at Photonics West has 100 spectral bands that range from 560 nm (green) to 1000 nm (near-IR), but different and wider wavebands are possible. The upper limit of a CMOS sensor’s waveband is limited to 1125 nm, which is the size of silicon’s bandgap. The lower limit depends on the choice of material and how it is modified. Lambrechts’ team is currently working on a new sensor that reaches 400 nm (violet).
The filter, which is made as a wafer, is fixed directly above the CMOS sensor to create a compact hyperspectral sensor that has no moving parts and which makes simultaneous use of all the light that falls on it. The spectral resolution of the IMEC camera is lower than that of a standard hyperspectral camera, but, thanks to its wide aperture, it has the offsetting advantage of high throughput and therefore fast operation.
At the IMEC booth in the Photonics West exhibition hall, Lambrechts’s colleague Bart Masschelein demonstrated the camera. He had it scan, in less than a second, over a collection of green objects: real leaves, life-like tissue leaves, and pieces of plastic. Masschelein’s computer interface quickly displayed the images and their corresponding spectra, which clearly distinguished the different materials. Masschelein’s software could also classify the objects based on their spectra—that is, knowing what the spectrum of a real leaf looks like, it could find the real leaves among the other objects.
Lambrechts hopes that the IMEC camera, being cheaper and more compact than standard hyperspectral imagers, will find applications beyond the traditional uses. But IMEC is a research center, not a manufacturer. If its cameras are put to work, say, inspecting vegetables on a conveyor belt or watching for poison gas on a battlefield, they will be built by IMEC’s industrial partners.