Array Camera Technologies at Fraunhofer IOF

Compact Multispectral Array-Camera

Simultanous recording of 66 parallel imaging channels with different transmission wavelengths.

Simultanous recording of 66 parallel imaging channels with different transmission wavelengths.

The particular challenge of multispectral cameras is the simultaneous detection of spectral and spatial information with high resolution. Scanning systems which measure either one spatial dimension or the spectral information in a time sequential mode are typically used. To overcome the associated problems, new developments have to operate in real-time or so called snap-shot mode. A second trend goes towards miniaturization, e.g. for applications such as  precision farming based on UAVs (unmanned aerial vehicle) for plant  monitoring.

Test scene and merged multi-spectral-image together with a pseudocolor image evaluating the NDVI.

Test scene and merged multi-spectral-image together with a pseudocolor image evaluating the NDVI.

In the project "MIRO", an innovative multispectral camera system was realized that combines both goals by using a multiaperture approach with 66 simultaneously working imaging channels each seeing its individual wavelength. The concept is based on a special microlens-array, which results in a very thin imaging optic of approximately 7 mm. For spectral selection, a linear variable bandpass filter is used. The definition of discrete spectral transmission bands for each channel is possible by aperture structures placed between the filter and the microlens-array. An additional crosstalk module is integrated to ensure the channel-separation on the image sensor. In conclusion, this system enables the recording of multispectral images ranging from 450 nm to 850 nm with a sampling of about 6.0 nm and a spatial resolution of 400 x 400 pixels.

The evaluation of the multispectral images is carried out by software developed in-house. Key features are the merging of sub-images and the extraction of local spectral information. Small imaging errors, as well as the angle  dependency of the used bandpass filter, are corrected. Furthermore, selected spectral channels can be combined for the calculation of classification indices such as the NDVI (normalized differences vegetation index) and shown as a pseudocolor image. The presented project is funded by the Fraunhofer Society within the framework of the research project "MIRO".

 

Authors: Robert Brüning, René Berlich, Christin Gassner, Martin Hubold, Robert Brunner