Hyperspectral imaging is a technique that adds a colorful third dimension to a reflected image that contains the target's spectral data. It can be used in applications such as topographic analysis of mineral deposits or farms, military surveillance, medical tissue analysis and archeological mapping. Hyperspectral imaging provides a wealth of light and composition data from imaging sensors in the field, in the lab and even in space.
Spectral imaging analyzes reflectance spectra, or light wavelength data. It might use technology such as reflecting mirrors, prisms, lenses and light sensors, much like the components and charge-coupled device (CCD) chips inside a digital camera. Combined with remote imaging technology, spectral imaging is used to measure wavelengths of the electromagnetic spectrum scattered by a target material. Devices called spectrometers and spectroradiometers note variations in the energy wavelength of the light reflected off a target and allow observers to determine compositional makeup of the material or landscape.
Hyperspectral imaging uses modern computing power to combine data from many images and add the third dimension of spectral data directly to the image. This data set is stacked into a “hyperspectral cube,” like a stack of snapshots, in which each pixel contains its spectral data. Multispectral imaging combines data of tens or hundreds of electromagnetic (EM) bands, but hyperspectral cubes can process data from thousands of bands.
Multispectral imaging normally utilizes data from multiple sensors, whereas hyperspectral data is often collected as a set of contiguous bands from a single sensor. The more data, the clearer the picture. The clearer the picture, the easier to determine from what substance or substances the subject is made.
Some applications of hyperspectral imaging include chemical analysis, fluorescence microscopy, thermal imaging, archaeological discovery and forensic investigation. Medical hyperspectral imaging extracts visual wavelengths of a spatial region and synthesizes the slices into a “topographical map” ready for clear medical analysis of tissue properties for various diagnoses or research purposes. This imaging technology can capture more of the EM band than visible light, including infrared and ultraviolet wavelengths, so it can enhance information that might otherwise go unseen by the naked eye. All materials contain spectral signatures that can provide vital clues for a plethora of applications across many fields.
For example, by understanding differences in chemical composition of soil and plant growth, forensic investigators are able to pinpoint otherwise unknown gravesites. This is because decomposition differentiates the reflectance spectra of plant growth from their surroundings. Put simply, the extra chlorophyll contained in plants fertilized by decomposition makes them stand out much more visibly in hyperspectral data than to the naked eye.
Remote sensing and digital imaging find new applications on an ongoing basis. Special libraries housing known spectral data of materials have been increasingly made available to researchers and civilians by organizations such as the United States' National Aeronautics and Space Administration (NASA). New applications for this technique have been continually developed in many industries. Agricultural uses might include determining plant varieties, water and nutrient conditions and the early detection of disease. As the technology becomes more available to the public, new applications are expected to be continually developed for great advantage over the relatively limited analytical power of single-point spectroscopy.
Thermal imaging technology has long been used in military or airborne surveillance. For this reason, special techniques designed to thwart this technology have been developed, in order to mask the heat signatures of ground forces from the air. Hyperspectral imaging might defeat these countermeasures with its multitude of spectral band measurements, offering precision analysis that can unearth the spectral “fingerprints” of the target.
The entire spectrum is gathered for each pixel of information, so the observer requires no prior knowledge of a material in order to make an analysis. Computer processing can include all available data for a complete analysis of a sample. This requires dedicated computing resources, including costly sensitive equipment and a large capacity of data storage. A hyperspectral cube represents multidimensional datasets requiring hundreds of megabytes each to process.