Hyperspectral and multispectral image processing

© Fraunhofer IOSB / M. Zentsch
Figure 1: Rapid food inspection - here hyperspectral imaging improves detection capabilities for automatic quality sorting or foreign object detection.

Hyperspectral Imaging

(HSI) combines the advantages of optical spectroscopy with those of spatially resolved image acquisition.

Conventional image processing systems with color cameras are very suitable for checking external quality features, such as the size, shape and color of a product. A particular challenge is the high variability of natural products, which makes it difficult to determine and test objective quality features. In food testing, internal quality parameters are also often important, such as the chemical composition or the freedom from internal defects of a product that cannot be externally detected. For this reason, spectroscopic methods that provide information about the chemical composition and physical properties of a product are suitable for evaluating internal quality parameters. HSI can literally make what is invisible to the human eye visible.

A hyperspectral image has a large number of spectral channels of closely spaced wavelength ranges, which can extend from the ultraviolet range to the long wave infrared. On the basis of the wavelength-dependent reflection behavior of a material, certain chemical properties can be measured, evaluated and graphically represented in a spatially resolved manner by HSI. One therefore speaks of chemical imaging.

This opens up a wide variety of fields of application for the automatic visual inspection. In particular for quality assurance in the food sector, HSI is successfully used for demanding tasks such as the detection of foreign bodies, optical quality sorting (see figure 1) or the inline inspection of agricultural products.

Figure 2: Hyperspectral data cube

Hyperspectral image processing

Hyperspectral images correspond to a data cube that has two spatial and one spectral dimension (see figure 2).

The three-dimensional data cube can be scanned in different ways by various optical recording techniques, with certain advantages and disadvantages occurring depending on the application. Line-scanning systems (e.g. when inspecting a material flow on a conveyor belt) and spectral-scanning systems (only recording static scenes) are used today in automatic visual inspection. Recently, special manufacturing processes have made it possible to scan a hyperspectral data cube instantaneously with a single image, which is also known as non-scanning or snapshot HSI. This technology enables hyperspectral image sequences to be recorded at a high frame rate and is therefore particularly suitable for inspections at high production rates.

Our offer

We build tailor-made, industrially applicable system solutions for your problem, which can be integrated into the existing systems.

Automatic visual inspection systems must almost always be selected and adapted individually and problem-specifically for the task at hand, as this is the only way to implement economical and robust solutions for industrial use. This applies in particular to HSI systems, which have a far higher degree of technical complexity than conventional visual inspection systems and whose use is associated with high investment costs. For most problems it is not clear in advance whether and how this relatively young technology can be used profitably. In contrast to the design of conventional camera systems, many parameters in HSI systems can only be determined experimentally.

This is why we at Fraunhofer IOSB have decided to always carry out an individual problem analysis as part of a technical preliminary investigation, which can provide information about the suitable spectral range, the number of required spectral channels and possible methods for hyperspectral data evaluation, among other things. This offers a target-oriented basis for decision-making for the selection of the optimal HSI technology. This may show that a task does not require a full HSI system, but that, for example, a color camera with an additional spectral channel in the near infrared is sufficient.

Projects and areas in which HSI is used:


Detection of pyrrolizindine alkaloids (PA) in plants

Joint project of Fraunhofer IOSB and the Julius Kühn Institute Berlin


Data fusion in the "FriDa - Fresh Data" project

The aim of the project is to reduce "food waste" with the help of NIR spectroscopy and additional sensors along food supply chains.



PROJECT "We save food"