Biologically-inspired data decorrelation for hyper-spectral imaging

Abstract

Hyper-spectral data allows the construction of more robust statistical models to sample the material properties than the standard tri-chromatic color representation. However, because of the large dimensionality and complexity of the hyper-spectral data, the extraction of robust features (image descriptors) is not a trivial issue. Thus, to facilitate efficient feature extraction, decorrelation techniques are commonly applied to reduce the dimensionality of the hyper-spectral data with the aim of generating compact and highly discriminative image descriptors. Current methodologies for data decorrelation such as principal component analysis (PCA), linear discriminant analysis (LDA), wavelet decomposition (WD), or band selection methods require complex and subjective training procedures and in addition the compressed spectral information is not directly related to the physical (spectral) characteristics associated with the analyzed materials. The major objective of this article is to introduce and evaluate a new data decorrelation methodology using an approach that closely emulates the human vision. The proposed data decorrelation scheme has been employed to optimally minimize the amount of redundant information contained in the highly correlated hyper-spectral bands and has been comprehensively evaluated in the context of non-ferrous material classification

Characterization of fine metal particles using hyperspectral imaging in automatic weee recycling systems

“…As observed in [6], the acquisition process of the hyperspectral image is highly affected by shadows, specular reflections (highlights), and inhomogeneous particles illumination. To compensate for these effects, we have tested two normalization methods, originally proposed in [7] and [8], as well as the Standard Normal Variate (SNV) method:…”

Automatic slag characterization based on hyperspectral image processing

“…These studies were undertaken with materials to be classified using only spectral information. The result was that the algorithms were insufficient to perform a robust classification, with obtained results of 56% [6].…”

“…This assertion was tested in earlier study [6]. These studies were undertaken with materials to be classified using only spectral information.…”

Characterization of Fine Metal Particles Derived from Shredded WEEE Using a Hyperspectral Image System: Preliminary Results

“…As observed in [8], beside heterogeneity due to the illumination system, the image acquisition process of metal particles in a laboratory is highly affected by shading effects, such as shadows and specular reflections (highlights). To compensate for these effects, two methods originally proposed in Stockman and Gevers [20] and Montoliu [21] were tested, along with the standard normal variate (SNV) algorithm: $R_{SG}\left(\lambda \right)=\frac{R\left(\lambda \right)}{true\sum _{i=1}^{N}R\left({\lambda }_i\right)}-\underset{j\in \left(1,,,N\right)}{min}\frac{R\left({\lambda }_j\right)}{true\sum _{i=1}^{N}R\left({\lambda }_i\right)}$ $R_M\left(\lambda \right)=\frac{R\left(\lambda \right)-{min}_{j\in \left(1,,,N\right)}R\left({\lambda }_j\right)}{true\sum _{i=1}^N\left(R,\left({\lambda }_i\right),-,{trueprefixmin}_{j\in \left[1,N\right]},R,\left({\lambda }_j\right)\right)}$ $R_{SNV}\left(\lambda \right)=\frac{R\left(\lambda \right)-{\mu }_R}{{\sigma }_R}$ where $R_{SG}$, $R_M$ and $R_{SNV}$ are the reflectance spectra computed with the methods proposed by Stokman and Gevers, Monotoliu and the SNV, respectively; N represents the number of spectral bands; ${\mu }_R$ and ${\sigma }_R$ are the average and standard deviation of the reflectance spectrum.…”