Poultry skin tumor detection in hyperspectral reflectance images by combining classifiers

Abstract

This paper presents a new method for detecting poultry skin tumors in hyperspectral reflectance images. We employ the principal component analysis (PCA), discrete wavelet transform (DWT), and kernel discriminant analysis (KDA) to extract the independent feature sets in hyperspectral reflectance image data. These features are individually classified by a linear classifier and their classification results are combined using product rule. The final classification result based on the proposed method shows the better performance in detecting tumors compared with previous works. © Springer-Verlag Berlin Heidelberg 2007.