Kapitel/Beitrag aus dem Buch: Längle, T et al. 2015. OCM 2015 – 2nd International Conference on Optical Characterization of Materials, March 18th – 19th, 2015, Karlsruhe, Germany : Conference Proceedings.
Hyperspectral sensors are becoming cheaper and more available to the public. It is reasonable to assume that in the near future they will become more and more ubiquitous. This gives rise to many interesting applications, for example identification of pharmaceutical products and classification of food stuffs. Such applications require a precise models of the underlying classes, but hand-crafting these models is not feasible. In this paper, we propose to instead learn the model from the data using machine learning techniques. We investigate the use of two popular methods: support vector machines and random forest classifiers. In contrast to similar approaches, we restrict ourselves to linear support vector machines. Furthermore, we train the classifiers by solving the primal, instead of dual optimization problem. Our experiments on a large dataset show that the support vector machine approach is superior to random forest in classification accuracy as well as training time.
Richter M. & Beyerer J. 2015. Towards many-class classification of materials based on their spectral fingerprints. In: Längle, T et al (eds.), OCM 2015 – 2nd International Conference on Optical Characterization of Materials, March 18th – 19th, 2015, Karlsruhe, Germany : Conference Proceedings. Karlsruhe: KIT Scientific Publishing. DOI: https://doi.org/10.58895/ksp/1000044906-8
This chapter distributed under the terms of the Creative Commons Attribution + ShareAlike 4.0 license. Copyright is retained by the author(s)
Dieses Buch ist Peer reviewed. Informationen dazu Hier finden Sie mehr Informationen zur wissenschaftlichen Qualitätssicherung der MAP-Publikationen.
Veröffentlicht am 18. März 2015