Spindle-Deviation Minimization Method for Classification in NDE System

Various feature extraction methods are developed in non-destructive evaluation (NDE). Among many linear methods been proposed, the Fisher linear discriminant is a very useful technique. But as the mean square error (MSE) increased, a classifier based on Fisher criterion will reduce the efficiency. In this paper, we expand the criterion and develop a spindle-deviation minimization (SDM) method to solve the classification problem. A comparison on global solution based on Fisher criterion with our method will be given, and it shows our method has good performance. Also, the results on our impact acoustic NDE system will verify the method efficient further.