Every material or compound interacts with light in their distinctive way. The material spectrum – i.e., how the amount of light reflected, emitted, or transmitted by the material varies over different wavelengths – depends on material chemistry and physical characteristics.

Hyperspectral imaging (HSI) combines spectral measurement with digital imaging. It is an increasingly used technique in industry, research, and remote sensing. HSI enables identification, mapping, and separation of materials by their chemical or color differences. As a quickly growing industrial machine vision technology, HSI allows industries to improve their processes, increase quality, and reduce waste.

Watch our tutorials to learn:


A line scan camera is based on a matrix detector and an imaging spectrograph. The incoming light passes through a high-performance fore optics and entrance slit, which creates the line imaging configuration of the camera. The line of light is dispersed by the spectrograph into a spectrum and detected by the matrix detector. As a result, one axis of the detector registers position from the line, and the other axis registers the spectral information in each spatial position in the line. This design offers considerable advantages:

  • The full spectrum for each spatial position is measured simultaneously, making the measurement insensitive to the movement of the sample or the camera itself.
  • Illumination is needed only for the narrow line, and much higher intensity can be obtained with lower total power than in 2D imaging configurations
  • Compared to filter based cameras, the imaging spectrograph collects light 5 to 15 times more efficiently and produces a purer spectrum.
  • As each frame from the matrix detector includes entire spectral and spatial data, results can be processed immediately after each line measurement.

Due to the above benefits, line scan cameras can be operated with high speed yet allowing them to obtain results in real-time with minimal processing. It makes line scan cameras optimal both for industrial and airborne environments. They can be easily integrated with an automated scanner stage to offer the highest performance and data quality in laboratory and field systems.

Some filter-based cameras, particularly those based on a linearly variable filer, are often improperly compared to line scan cameras even though they collect data of a 2D area. The filter-based cameras have another severe limitation in addition to the lower light collection efficiency as they collect spectral bands from different spatial locations. It leads to co-registration challenges and much more complicated data processing than with line scan data.

More information about line scan cameras and how they compare to other HSI technologies can be found in:


Where can hyperspectral cameras be used, and what information can they provide? What does it mean to integrate HSI in the industrial production process, and what parameters need to be considered?


Hyperspectral cameras are sensitive to specific wavelengths, and the spectral range selected should be based on the measured material and its characteristics. There is plenty of scientific research done with hyperspectral cameras if you have no prior knowledge of how spectral features (signatures) are related to the material characteristics.

On our product page, you find a good overview of applications that can be solved with each product and wavelength range they cover.

See also our short video to learn how to select the correct spectral camera for the application.


When selecting the illumination for a spectral imaging system, three major points must be considered.

Illumination Strength

Spectral imaging requires much more light than traditional RGB or grey scale black-and-white imaging. The required total illumination power depends on

  • the geometry of the illuminating beam and the distance from the sample,
  • width of the illuminated target (like conveyer belt)
  • the integration time per the image

Need for illumination power increases with

  • longer distance to the target
  • less illuminating light collimated to the target
  • shorter integration time.

The wavelength range of the illumination source

Illumination for hyperspectral imaging must always have a continuous spectrum that covers the full wavelength range of the hyperspectral camera. If the illumination does not have a signal in some of the wavelengths, the measured data on those wavelengths is not valid.
The illumination types compatible with hyperspectral imaging sensors are

  • For visible range, most typical is halogen illumination (spots or linear), LED is also used. Supercontinuum lasers show good potential.
  • For NIR range: halogen is most typical. LEDs and supercontinuum lasers show good potential.
  • For SWIR range: halogen illumination
  • For MWIR and LWIR range: thermal illumination

Illumination uniformity

It is important to ensure that the illumination covers the full measurement area with uniform intensity and spectral range, minimum shadows, or specular reflections. To learn more about the illumination for spectral cameras, watch our tutorial video on how to estimate required illumination power for a hyperspectral camera (applies to halogen spot based illumination).


Most industrial sorting and quality inspection applications require a high image rate to accurately monitor small objects traveling at fast-line speeds. A higher image rate limits the integration time (the time that the camera has to collect light reflected from the material stream). Specim FX cameras provide both the highest image rates and light collection efficiency in the market to keep up with the industrial speed requirements.

How to select the correct frame rate for your application:


In addition to the spectral camera and illumination, specific data processing is required for a complete application solution. Knowledge of how the object spectrum is affected by the object characteristics (e.g., by its chemical composition) is needed to build classification and sorting models, which are then applied in real-time to the data acquired by a spectral camera.
The hyperspectral data is an enormous source of information and requires adequate processing power from the system. The data contain repetitive information where the actual chemical information of the material is often carried by a limited spectral range or number of feature wavelength regions. Traditional so-called chemometrics methods and rapidly developing machine learning techniques are very powerful to extract the relevant information from this type of data. They apply mathematical and statistical methods and are often used to obtain classification or quantifications results from the data measured by a hyperspectral camera.

There are several commercial software solutions available that provide the required tools to build classification and prediction models for hyperspectral data. The suppliers that support Specim cameras are CAMO, Perception Park, PerClass Mira, and LuxFlux. These software give output in standard machine vision formats; thus, integration to existing machine vision software (e.g., Halcon, Cognex, Labview, or Sherlock) is straightforward. The results can then be used to control the air nozzles or picking robots.


Specim FX series cameras provide the speed and flexibility required by industrial processes. Camera speed can be adjusted based on the application requirements. The tolerance for delays of the entire inspection chain, i.e., from acquiring the spectral image to taking action by system control, is typically very strict. Specim FX series cameras are designed to meet these requirements. An industrial PC or another high-performance platform is usually required to run the classification in real-time.