Hyperspectral Imaging - Fraunhofer ITWM

Enhanced Color Perception: From Red, Green, Blue to Hundreds of Color Channels

Hyperspectral imaging is revolutionizing the way we perceive color and light. Normal cameras capture images in three colors - red, green and blue. This limitation is sufficient for many applications, but hyperspectral cameras go far beyond this. They use several hundred color channels and thus enable extended color perception. This is particularly useful for scientific and industrial applications where accurate spectral information is required.

Thanks to the many color channels, hyperspectral cameras can precisely capture the spectral properties of samples. This means that we can use them to detect differences in materials and substances that remain invisible to the human eye or conventional cameras. This technology is used, for example, in materials testing, agriculture, environmental monitoring and many other areas.

We use hyperspectral cameras for visible light and near-infrared (non-visible) light. This expands the application, as the near-infrared spectrum provides information that is not available in the visible range. This makes it possible, for example, to visualize the water content in plants or certain chemical compositions in materials.

Structure of a Measurement System - Camera and Lens: The Heart of Hyperspectral Imaging

A hyperspectral measurement system consists of a camera equipped with a special lens. This camera is able to capture the numerous color channels and thus provide detailed spectral data. The lens must make optimum use of the lighting conditions and enable precise focusing of the sample.

Homogeneous Lighting: The Key to Precise Data Acquisition

In addition to the camera, uniform and homogeneous illumination of the sample is crucial (usually in the form of a tube, not as in the image below). Without optimal illumination, the spectral data can be distorted or inaccurate. It is therefore important to use an illumination source that ensures the most uniform illumination possible. This can be achieved using special lamps or lighting systems designed for hyperspectral applications.

With this equipment and knowledge of hyperspectral imaging, we can perform detailed and precise examinations that go far beyond the capabilities of conventional cameras. Whether in research, industry or environmental monitoring, hyperspectral imaging opens up new dimensions of analysis and diagnostics.

Our hyperspectral cameras are equipped with a 2D detector, similar to normal cameras. The difference, however, lies in the use of dimensions: One dimension is used for the spectral information, the other for a line on the sample. This means that we measure a line on the sample and obtain an image in 2D (see image).

Efficient Product Measurement on the Conveyor Belt

The method is very suitable for measuring samples and products on a conveyor belt: The products pass under the camera on the conveyor belt. There, the samples are illuminated evenly and an image is taken with the width of the conveyor belt. If the images are generated fast enough, the products are measured as they move. The cameras generate the images very quickly – typically in the region of more than 100 images per second.

Evaluation: Precise Real-Time Evaluation Using Artificial Intelligence

We generally use Machine Learning (ML) methods for evaluation. Depending on the application, there are various ML methods to choose from. The evaluation takes place in real time and the software is individually adapted to the needs of the company or application.

© Fraunhofer ITWM
Efficient product measurement on the assembly line with hyperspectral imaging

© freepik / Fraunhofer ITWM
The example shows the changed spectral response of an avocado during the ripening process.

Diverse Applications of Spectral Analysis

The most common applications are focal points based on spectral information. This means distinguishing products based on different spectral characteristics as in the following examples:

Finding foreign bodies in bulk materials (which can also look the same to the human eye): such as white plastic in paper, metal in plastic granules, etc.
Sorting products: separating different plastics, sorting waste, sorting coated or uncoated products
Recognizing changes in products: for example, the degree of ripeness of fruit or vegetables, changes in the concentration of solutions, etc.

Precise Measurement of Coating Thicknesses Using Broadband Spectral Analysis

Because so many colors are detected and the measuring system has such a wide bandwidth, we can also measure the layer thickness of thin samples (e.g. thin films). The light is reflected at each boundary layer. This causes interference on the camera. The thinner the layer, the longer the wavelength of the interference oscillations. For thin layers, you need a lot of bandwidth (i.e. many color channels over a large wavelength range).

We can measure films and other thin layers between approx. 2 µm and 100 µm. The measurements are very fast and have a good repeatability of better than 1 µm.

© Fraunhofer ITWM
Measurement of a 3 µm thick film (160,000 measuring points measured in approx. 2 s)

Example Projects

Hyperspectral Imaging for Ripe Tomatoes

Using hyperspectral imaging, we determine the BRIX or sugar concentration of tomatoes, which allows conclusions to be drawn about the degree of ripeness.

ASKIVIT

In the project »ASKIVIT«, we are working with partners on automated sorting to recover wood, wood-based materials and non-ferrous metals. Our experience in the field of terahertz imaging, among other things, is supporting here.

Hyperspectral Imaging – More Than Just Red, Green and Blue