Hyperspectral imaging can be considered a paradigm shift in image processing systems. Instead of the three colour channels common in artificial vision, hyperspectral imaging uses up to hundreds of channels. This makes it possible to detect very subtle differences. In addition, hyperspectral cameras often have an extended spectral range that goes beyond the visible, i.e. into the short-wave infrared light spectrum (SWIR spectrum).
Thus, imaging enables the determination of chemical and physical properties of a product, e.g. checking the quality of organic products in the food industry. The Spanish company IRIS Technology has developed solutions to use hyperspectral imaging for inline industrial inspection. This includes suitable hardware and machine learning algorithms tailored to the specific application. An example of a successful implementation of hyperspectral imaging in an industrial application is the solution developed for a large industrial bakery to determine the fat content of biscuits and pastries in real time and continuously.
In the food industry, it is essential to keep the manufacturing process under control. This ensures that the quality and taste of the product remain constant for a given recipe. Controlling the fat content is particularly important. Moreover, large fluctuations in fat value lead to cost overruns due to sub-optimal use of oil as a raw material, as well as unexpected changes in the palatability of the product. IRIS' client, who specialises in biscuits and pastries, observed changes in the product that could hardly be explained by recipe changes. An investigation was therefore launched to find out which processes were causing these changes.
The investigation was lengthy and complicated because current laboratory techniques for checking fat levels are carried out offline and require special sampling, preparation, equipment and personnel. It takes several hours to get the results and it is not possible to correct the process in real time. This makes these methods incompatible with the idea of standardising the product.
What the factory needed was a continuous measurement with a minimal margin of error to quickly measure fat content and relate it to changes in its processes. The company turned to IRIS to implement an inline hyperspectral imaging solution that performs a 100% inspection of the products and determines their fat content in real time. The system is based on the IRIS VISUM HSI hardware platform. VISUM HSI is based on a line scan camera with SWIR sensitivity. This makes it well suited for fat content measurement. The system can capture up to 300 lines per second to inspect products without slowing down the process. The built-in software and chemometrics tools allow the user to self-calibrate the instrument as product composition changes, and it integrates seamlessly with plant information systems.
