Hyperspectral imaging (HSI) is an advanced technique used for capturing and analyzing a wide spectrum of light beyond what the human eye can see. This technique divides light into many different spectral bands and captures detailed information about each band. It’s widely used in fields like remote sensing, environmental monitoring, agriculture, mineralogy, and medical diagnostics. HSI techniques can be mainly categorized into three types: spectral scanning, spatial scanning, and snapshot imaging. Note: the following is not an exhaustive survey; some techniques such as Fourier Transform (FT) spectroscopy are not detailed. Also not covered are methods in which the sample is illuminated by sequential band-passes of light.
Each method has unique characteristics and applications.
Overview and High Level Comparison
| Feature | Spectral Scanning | Spatial Scanning [Snapscan | Pushbroom] | Snapshot Imaging |
|---|---|---|---|
| Speed | Slower (one wavelength at a time) | Moderate (line by line) | Fast (entire image at once) |
| Complexity | High (requires precise wavelength tuning) | Moderate (involves mechanical scanning) | Low (simple capture mechanism) |
| Resolution | High spectral resolution | High spatial resolution | Balanced spatial and spectral resolution |
| Applications | Precision wavelength analysis, laboratory use | Remote sensing, quality control, environmental monitoring | Real-time applications, industrial sorting |
Having provided an overview and high level comparison of the different methods, in the following sections, we will describe the different techniques in a manner that details their specific strengths.
The rest of this article is about Spectral Scanning Techniques. The following articles describe Spatial Scanning and Snapshot Imaging techniques.
Spectral Scanning Techniques
Spectral scanning methods acquire images by sequentially imaging the entire field-of-view, capturing one spectral band at a time.
The time required per image is usually quite short; the time needed to acquire a spectral data cube depends on how fast the system can change from one wavelength to another. Key technologies in this category include:
- Using Multiple Filters, deployed on a Filter Wheel or slider: This may be considered the “conventional” method of obtaining several spectral images of a given field-of-view. An electromechanical system positions a bandpass filter in the optical path allowing a single image to be captured for that particular band. A sequence of images may then be acquired, one for each of “N” bandpass filters. The images are stacked in software to create the N-plane spectral data cube.
- Liquid Crystal Tunable Filters (LCTF): LCTFs use electronically controlled liquid crystal elements to selectively pass light of specific wavelengths. They offer high spectral resolution and are suitable for applications requiring precise wavelength selection.
- Acousto-Optic Tunable Filters (AOTF): AOTFs use sound waves to diffract and filter light into different wavelengths. They are known for their fast tuning speed and are used in applications requiring rapid wavelength switching.
- Micro-Electro-Mechanical Systems (MEMS) Based HSI systems: A MEMS-based method is used to vary the effective thickness of an etalon [formed via two parallel, flat, semi-transparent mirrors] that is placed in front of the image sensor. This forms a tunable Fabry–Pérot interferometer in front of the pixel array. “Tuning” the distance between the two parallel surfaces changes the wavelengths of light that are transmitted through the etalon. For a given field-of-view a sequence of band-pass images are collected, thus forming a hyperspectral datacube.
