'ON THE FLY' DIMENSIONALITY REDUCTION FOR HYPERSPECTRAL IMAGE ACQUISITION

Hyperspectral imaging (HSI) devices produce 3-D hypercubes of a spatial scene in hundreds of different spectral bands, generating large data sets which allow accurate data processing to be implemented. However, the large dimensionality of hypercubes leads to subsequent implementation of dimensionality reduction techniques such as principal component analysis (PCA), where the covariance matrix is constructed in order to perform such analysis. In this paper. we describe how the covariance matrix of an HSI hypercube can be computed in real time 'on the fly' during the data acquisition process. This offers great potential for HSI embedded devices to provide not only conventional HSI data but also preprocessed information.