Adaptive Spatio-Spectral Hyperspectral Image Processing for Online Industrial Classification of Inhomogeneous Materials

An approach for considering spatio-spectral information when classifying inhomogeneous materials in industrial environments is proposed. Its main application would be in the inspection and quality control tasks. They system core is an ANN based hyperspectral processing unit able to perform the online determination of the quality of the material based on its composition and grain size. A training adviser is being implemented in the system in order to automate the determination of the optimal spatial window size, as well as to reduce the number of spectral bands used and for determining the optimal spectral combination function through the automatic extraction of the discriminating features. Several tests have been carried out on synthetic and real data sets. In particular, the proposed approach is used to discriminate samples of andalusite having different purities; the results obtained show an accuracy of better than 98%