HSI cameras in 2024, a quick guide
Hyperspectral imaging (HSI) cameras break down the spectrum into numerous narrow bands, enabling the detailed analysis of materials based on their spectral signatures. This technology provides a comprehensive understanding of the composition and characteristics of the imaged objects or surfaces.
Size and weight are largely dependent on the optical arrangement used to separate the spectral bands. While some HSI cameras tend to be much bigger and heavier than standard RGB cameras, due to bulkier optical systems with prisms and gratings, and even mechanical scanners; there are also systems with sensor-integrated filters that are as compact and light as their RGB counterparts. Like in traditional cameras, the cost of a HSI camera depends on two main factors: the optics and the sensor. Bulky optics around a prism or grating are more expensive than sensor-integrated filters. As well, sensors in the SWIR range like InGaAs are much more expensive than CMOS sensors in the visible-NIR spectrum.
Hyperspectral data acquisition can occur through two primary methods: snapshot (e.g. PhotonFocus spectral cameras) or scanning (e.g. Specim spectral cameras). Snapshot acquisition is characteristic of sensor-integrated cameras and captures the entire hyperspectral cube in one instant, suitable for dynamic scenes. Scanning is characteristic of HSI cameras with optics based on a prism or grating and involves sequentially capturing spectral bands, offering better spectral performance.
While sensor-integrated filter cameras feature reduced cost, weight and size as well as snapshot and realtime acquisition capabilities that improve usability, they also involve a loss in terms of spectral resolution, spatial resolution, or both.
Scanning HSI cameras may be bulkier and slower, but they bring exceptional spectral and spatial resolution, allowing for the capture of detailed information in both the spectral and spatial domains. High spectral resolution enables the differentiation of subtle variations in material composition, while spatial resolution adds to the clarity of images. This capability makes HSI scanning cameras a requirement for some tasks requiring a high precision in analysis and identification.
HSI cameras generate datacubes, three-dimensional arrays where each pixel contains spectral information, allowing users to extract meaningful insights and make informed decisions in fields ranging from agriculture and environmental monitoring to healthcare and beyond. However efficient datacube acquisition and processing can be challenging. To start with, the need of datacube reconstruction and sometimes calibration makes putting a HSI camera to work much harder compared to a standard RGB camera. And this is generally true for any type of HSI camera.
But difficulties don’t stop here. As our eyes can only see trichromatic representations, visualization and analysis of HSI data is not straightforward. Tailored processing and ML approaches are needed that cater to the additional dimensionality of the data. As if that weren't enough, datacubes are usually much heavier than RGB images.
All of this hampers the development and deployment of solutions based on hyperspectral cameras. In this regard, SpectralEdge brings a cost effective solution that provides state of the art hardware and software tools together with a cloud API to greatly smooth the test, development and deployment of solutions that benefit from the insights delivered by HSI cameras.