Hyperspectral target identification using near IR supercontinuum light source


Reflected light in the infrared spectral region provides fingerprint identification information for targets ranging from vegetation to minerals and anthropogenic materials. Broadband eye-safe infrared illumination allows this Light Detection and Ranging (lidar) technique to be operated during the day and at night. Our approach relies on a developed supercontinuum light source providing bright, broadband illumination with laser-like directionality. As the result, a record-breaking measurement range of 1.5 km was demonstrated.

Hyperspectral detection is a widely used technique for identifying materials and their properties. The technique is based on the tendency of different substances to absorb and reflect light differently. This property makes distinguishing materials possible based on spectral response. The variability of measured targets dictates the required spectral range and number of detection bands. For example, it takes only two bands to separate apples from tomatoes, whereas additional bands are needed to separate apple varieties. The human eye has only three colour receptors: red, green and blue, whereas the infrared spectrometer used in this study measures up to 256 spectral bands to discriminate between targets.

A supercontinuum (SC) light source was developed and used as an active illumination. SC is an ideal source for remote hyperspectral sensing because of its laser-like directionality, broad wavelength coverage and high spectral power density. Such combination allows for active illumination and measurement of reflection spectra from diffusive targets over distances exceeding one kilometre.

The instrument has shown feasibility for automated real-time target identification, using both indoor and in the field measurements. Four different identification algorithms are implemented for target recognition. Optimization of the algorithms is however application specific. For example, some algorithms can reliably distinguish numerous different materials, whereas other algorithms may better track small composition variations of pulp.

The lidar technology is versatile and quite general; the same approach can be used for long-range detection of military targets, mineral exploration, and vegetation inventory studies, whereas with small modifications the technique is applicable to waste sorting or on-line industrial quality control. Current development is focused on miniaturization of optics, implementation of an improved, low-cost detection scheme, optimizing of measurement speed and further algorithm development. A detailed description of the instrument can be found at: Manninen et al Optics Express 22, 7172-7177 (2014).

hyperspectral 2.jpg