20.5. Kernel methods for estimation and classification of data from spectral imaging

Public examination of a doctoral dissertation in the field of computer science

Doctoral candidate: M.Sc. Ville Heikkinen 

Date and venue: 20.5.2011 at 12 noon, Louhela Auditorium, Joensuu Science Park 


In this dissertation Ville Heikkinen has studied color calibration of a camera, estimation of reflectance spectra, and supervised classification of spectral information, based on reproducing kernel Hilbert space methods (RKHS). A unifying characteristic of our spectral data was that imaging was performed with small number of broad-band spectral response functions. 

We considered reflectance estimation as a generalized color calibration problem and mainly focused on an empirical regression approach that assumes relatively large ensembles of training data. The connections of several reflectance estimation and color calibration models to more general RKHS models are discussed. Several RKHS models and transformations based on physical a priori knowledge are introduced and evaluated for the reflectance estimation from responses of an ordinary RGB camera. 

The results suggest that new models lead to better accuracy in reflectance estimation and color calibration than some classical, more widely used models. The data classification is discussed in remote sensing context, where data are simulated to correspond measurements from a multispectral airborne camera. In the classification of birch (Betula pubescens Ehrh., Betula pendula Roth), pine (Pinus sylvestris L.) and spruce (Picea abies (L.) H. Karst.) trees, a Support Vector Machine classifier (SVM) and RKHS feature space mappings were used to validate the performance of several simulated sensor systems. 

The results indicate a need for careful data pre-processing, a higher number of sensor bands, decrease in the bandwidths or new positioning of the bands in order to improve pixel-based classification accuracy for these tree species. 

The doctoral dissertation of Ville Heikkinen entitled “Kernel methods for estimation and classification of data from spectral imaging” will be examined at the Faculty of Science and Forestry. The opponent in the public examination is Professor Erkki Oja, Aalto University and the custos is Professor Markku Hauta-Kasari of the University of Eastern Finland.

Photo available for download at http://www.uef.fi/vaitoskuvat

Contact: Ville Heikkinen, tel +358 50 3496113, ville.heikkinen@uef.fi

Publishing year: 2011

Back to this years article listing