Analysis and classification of hardwood species based on Coiflet DWT feature extraction and WEKA workbench

The work proposes to introduce Coiflet discrete wavelet transform (DWT) family, to extract features of microscopic images of hardwood species in order to classify them into 25 different hardwood species. The images are being decomposed into 3 levels using Coiflet DWT family. Overall 48 features are obtained for each of the images with mean, standard deviation, kurtosis and skewness extracted from each of the 12 subimages. Images of hardwood species have been classified by a pertinent application WEKA 3.7.9. Several WEKA classification algorithms have been tested on 48×500 feature matrix generated by the Coiflet DWT family, and it is found that multilayer perceptron classification algorithm belonging to function category of WEKA give classification accuracy of 92.20% for the feature matrix produced by “coif2” discrete wavelet transform. The same amount of accuracy is also obtained for the features extracted by “coif1” DWT, using logistic classification algorithm.