From Hyperspectral Edge Computing to Smart Imaging

For the manufacturing plants of tomorrow, it is important to be able to display an increasing number of different materials in a single product stream and to make key data available in real time. Many hyperspectral systems compromise an edge computing engine for fast data reduction. To extract key information and get the most out of a broad bandwidth hyperspectral data stream, classification and regression algorithms must be developed in advance.

This presentation shows the results of a closed hyperspectral data processing loop, from industrial data acquisition via machine learning based algorithm training to real-time feature extraction.

About the presenter
Matthias Kerschhaggl, Ph.D. is CTO and owner of EVK, an Austrian based expert company for industrial imaging. He is engaged in data science and analytics, predominantly dealing with data streams stemming from sensor-based sorting and control applications used in industries such as food, chemical, mining, and pharmaceuticals. He holds a doctorate in experimental physics and has more than 15 years of experience in the fields of statistical learning and data mining of datasets from various areas including, astroparticle physics, integral field spectrographs, and hyperspectral and inductive imaging.