Hyperspectral image processing is playing an increasingly important role in the food industry. Unlike other quality-control methods, it can be used to monitor the chemical composition of a food product in real time during processing. However, interpreting the output of hyperspectral cameras is a much more complex and less intuitive task than processing data from other quality control devices such as colour cameras. Therefore, using hyperspectral technology successfully depends on having an intuitive and easy-to-use data processing system. That is what BT-Anlagenbau is offering with the HSP Food Analyzer.
The HSP Food Analyzer makes it easy and economical for engineers to integrate hyperspectral cameras into their process systems. They only have to connect the high-performance industrial computer with the preinstalled HSP software package to their network and install a suitable hyperspectral camera. An easy-to-use software protocol allows any control system to communicate with the analyzer and use the data it supplies.
Every HSP software package includes the HSP Configurator, which allows the user to configure the necessary recognition function easily and intuitively – without any expert help. The objects that need to be distinguished from each other are imaged, the differences are marked and then the recognition function is ready to use. The user interface of the HSP Configurator enables users to use the algorithms for spectral analysis by point and click, intuitively and without deeper knowledge. The users can check the outcome of their settings quickly and easily by means of false-colour pictures without having to analyze spectral signals.
The HSP Food Analyzer converts the spectral signals from the camera to I/O data in real time. The data can then be read and processed by a higher-level system (e.g. a PLC).
Every pixel from a hyperspectral camera consists of a spectral signal which in turn is made up of many hundreds of values. The HSP Food Analyzer converts this signal in real time into one that is suitable as an input for a control system. The signal that is sent to the control system is usually reduced to the relevant quality result for each pixel … usually a simple OK/not OK.
In contrast to normal colour camera systems which register just three colour points per pixel (red, green and blue), hyperspectral cameras detect hundreds of points of spectral data per pixel. These data can be displayed as a spectral signal. Analyzing this signal by means of suitable mathematical algorithms, so-called chemometric methods, makes it possible to extract information on the chemical composition of the imaged object.
Food in General
Fruit & Vegetables
Recognition of overripe and spoiled spots
Measurement of fat; detection of parts with high fat content
Detection of pits, pit fragments and foreign objects
Measurement of ripeness and freshness