Principal component analysis for poultry tumor inspection using hyperspectral fluorescence imaging

This paper presents detection of skin tumor on poultry carcasses using hyperspectral fluorescence images. Image samples are obtained from a hyperspectral imaging system that provides digital images of 65 spectral bands with wavelength ranging from 425 [nm] to 711 [nm]. The principal component analysis (PCA) technique finds an effective representation of spectral signature in a reduced dimensional feature space. A support vector machine (SVM) classifies the feature vectors and makes a decision whether each pixel falls in normal or tumor categories.