Application of kernel-genetic algorithm as nonlinear feature selection in tropical wood species recognition system

► KDA with GA are used to generate nonlinear features of the wood features. ► Use dataset of 52 wood species, each has 70 images for training and 30 for testing. ► 157 wood features are extracted from the fusion of BGLAM and SPPD methods. ► Size of original training database is reduced by 61.75% using the proposed method. ► Accuracy of 98.6%, reduction of 3 times the memory space for storing the samples.