This page summarizes possibilities on applying chemometric models from 3rd party software onto hyperspectral data for off-line or in-line purposes.
Chemometric modelling is a very essential step when investigating to chemically or physically material properties. Typically, experts like to overcome this job with their familiar environment (software).
This natural behavior is supported by the export of data (spectra, cubes) and by the back import of models into the Perception Studio software program. When configuring the real-time processing core (Perception Core) by the Perception Studio, a data processing system is obtained, allowing to do multiples of predictions per spatial object pixel in industrial real-time.
When doing predictions in the industrial environment, a number of data processing steps must be considered:
The sensor data processing aims to a "standardization" of the instrument.
Therefore electrically and optically disorder (like sensor noise, optically distortions, etc.) are corrected by this data processing step. Data from different instruments (same type but different batch) ought to get comparable to each other.
The hyperspectral preprocessing aims a form of spectra best suited for further investigation like analysis.
The feature extraction aims the extraction of a certain value out from spectroscopic data. This value can be the prediction value obtained by applying a linear model onto spectral data. Since this step is typically done in parallel, per spatial object pixel multiple predictions can be obtained simultaneously.
Sometimes also feature operation is an important step. By feature operation a number of features (output of the feature extraction), like predictions, are mathematically combined with each other.
Applying models from 3rd party software is possible without any restrictions, when
Preprocessed means: sensor data preprocessing as well as hyperspectral preprocessing was applied to the spectra using the Perception Studio program (not the 3rd party software).
A linear model is described by the vector B and the scalar B0. X is a matrix of spectra (input) and Y is a vector of predictors (output).
The typically workflow on getting and applying a linear model is:
In case of real-time processing:
© 2019 by Perception Park GmbH
The content of this page and any attached files are confidential and intended solely for the addressee(s). Any publication, transmission or other use of the information by a person or entity other than the intended addressee is prohibited. If you receive this in error please contact Perception Park and delete copied material. Perception Park GmbH, Wartingergasse 42, A-8010 Graz; Austria; FN 400381x