Hyperspectral imaging is coming of age
Hyperspectral imaging has not been used extensively in industrial applications due to the complex analysis it required. The technique makes use of the fact that organic materials selectively absorb light at different wavelengths in the infrared region of the spectrum depending on their composition. This results in distinctive "fingerprints" that can be used to identify them.
In hyperspectral imaging, a series of images are built by sequentially allowing narrow wavelength bands of infrared (IR) light from the sample to fall on the sensor. These images are combined to form a three-dimensional hyperspectral data cube. This data cube contains all the information needed to extract the chemical composition at each pixel. The vast amount of information usually required a chemist or a mathematician to crunch the data and analyze the results. The complex process was so costly that it was only used in a handful of applications such as waste sorting for recycling.
A new approach
Now the technique has evolved with the emergence of affordable, flexible, high-speed data processing software. A spectrograph is combined with a camera that is sensitive to IR wavelengths. The spectrograph allows the reflected light from the sample to be sorted into its constituent wavelengths. This approach is called "chemical color imaging" and makes chemical material properties accessible to the machine vision engineer.
The software extracts data from the complex data cube, which is processed in real time to produce an image where the output of each pixel is color-coded according to the chemical composition of the material at which it is looking. This is invaluable because the system reveals information that other machine vision technologies cannot.
Rather than processing data from the entire spectrum, it may be possible to tune into just a few key wavelengths that would distinguish between specific materials or identify known contaminants that could arise from a production processing stage. By reducing the data processing, the image acquisition speed can be increased while reducing the inspection time.
The data processing software is designed to turn the camera system into an intuitive configurable chemical color camera. It provides a user interface that defines the acquisition parameters and acquires the initial image data. From there, the hyperspectral data is explored with images and graphical feedback to help define the best models for extracting the key information. Then it transfers the color-coded images into a standardized machine vision form. Online processing using image process methods such as gray level analysis and blob detection can be used for color sorting in machine vision applications.
Chemical color imaging in action
Another benefit of hyperspectral imaging is that a lot of packaging material is transparent to the IR light meaning that the technique can be used to examine products inside the packaging. The technique has potential applications in the food and pharmaceutical industries, where cross-contamination in a packing line, for example, could have potentially life-threatening consequences for consumers, and related implications for the manufacturer also could be massive, in terms of reputation and costly product recalls, and possible production line closure while the problem is investigated.
Rob Webb is a technical specialist at Stemmer Imaging. This appeared on the Control Engineering Europe site on Nov. 28, 2016, and was edited for the Control Engineering international page by Chris Vavra, production editor, CFE Media, Control Engineering, firstname.lastname@example.org.
- Hyperspectral imaging is done when images are built by sequentially allowing narrow wavelength bands of infrared (IR) light to fall on the sensor.
- High-speed data processing software has been developed to make hyperspectral imaging more affordable.
- Hyperspectral imaging can be used to examine products inside the packaging for applications that require safety.
What other applications can hyperspectral imaging be used for?
See the original article here.
See other Control Engineering international coverage here.