A SVM classifier combined with PCA for ultrasonic crack size classification

Pattern recognition may be used for crack size and type classification in ultrasonic nondestructive evaluation. Feature selection and reduction of computational complexity are two important problems to be solved in the development of pattern recognition algorithms. This paper describes a classifier based on support vector machines (SVM) and principal component analysis (PCA). The proposed approach can reduce the dimension of the feature vector by using PCA, which can dramatically reduce the input data dimension for SVM classification. The kernel fisher discriminant (KFD) is also described, which helps to select the parameters of the kernel function in SVM. Classification results using experiment data show the effectiveness of the proposed approach.