A Recognition Method of Fracture Images of Aerial Material Based on Grouplet Entropy-RVM

Grouplet transform is a multi-scale analysis technology of two-dimensional image through Haar transform, which has the ability to adaptively change the basis when the image texture has changed. Compared with the wavelet transform, Grouplet transform has better ability in sparse representation and it has more superior performance in to deal with the complicated metal fracture of texture image recognition. Proposed a new recognition method based on Grouplet entropy and relevance vector machine. In the proposed method, Grouplet entropy is used as the image's feature, RVM as a classifier. The experimental results show that this method combined the advantages of Grouplet and RVM. The proposed method get 85.58% of correct recognition rate for 222 metal fracture images. Compared with Grouplet entropy-SVM, the recognition speed of Grouplet entropy-RVM is about five times than Grouplet entropy-SVM.