Classification of material types using laser speckle, wavelets and artificial neural networks

In this paper, a method for classifying material types is proposed where wavelets and neural networks are applied on laser speckle images. Multiresolution wavelet analysis was used for feature extraction while neural networks were used for classification. To improve the classifier performance, a reduced set of wavelet coefficients was obtained from all levels of transform resolutions. The effect of the laser angle of incidence on classification sensitivity was investigated. It was found that maximum sensitivity was achieved at an angle of 45 degrees. Results obtained showed that the proposed method was capable of classifying eight different materials from similar and different brands.