Semi-automatic supervised classification of minerals from x-ray mapping images

This paper addresses the problem of assessing the robustness with respect to change in parameters of an integrated training and classification routine for minerals commonly encountered in siliciclastic or carbonate rocks. Twelve chemical elements are mapped from thin sections by energy dispersive spectroscopy (EDS) in a scanning electron microscope (SEM). Extensions to traditional multivariate statistical methods are applied to perform the classification. Training sets are grown from one or a few seed points by a method that ensures spatial and spectral closeness of observations. Spectral closeness is obtained by excluding observations that have high Mahalanobis distances to the training class mean. Spatial closeness is obtained by requiring connectivity. The marginal effects of changes in the parameters that are input to the seed growing algorithm are evaluated. Initially, the seed is expanded to a small area in order to allow for the estimation of a variance-covariance matrix. This expansion is controlled by upper limits for the spatial and Euclidean spectral distances from the seed point. Second, after this initial expansion the growing of the training set is controlled by an upper limit for the Mahalanobis distance to the current estimate of the class centre. Also, the estimates of class centres and covariance matrices may be continuously updated or the initial estimates may be used. Finally, the effect of the operator's choice of seed among a number of potential seeding points is evaluated. After training, a standard quadratic classifier is applied. The performance for each parameter setting is measured by the overall misclassification rate on an independently generated validation set. The classification method is presently used as a routine petrographical analysis method at Norsk Hydro Research Centre.