New Signal Processing Scheme for the Analysis of Electromagnetic Images
Electromagnetic methods of nondestructive testing find widespread application in industry. A vast majority of the defect characterization schemes using electromagnetic methods involve estimation of the size and/or shape of the defect on the basis of a one dimensional signal obtained by scanning the surface of the test specimen using a suitable transducer [1–3]. Recent years have witnessed increasing interest in the development of imaging techniques for characterizing defects. As an example, eddy current imaging methods involve a raster scan of the surface of the test specimen to obtain a two dimensional image whose elements represent the real or imaginary components or alternatively the magnitude or phase of the impedance of the eddy current probe [4,5]. In the case of magnetostatic imaging methods, the specimen under test is scanned by a flux sensitive transducer such as a Hall probe. The image is obtained, typically, by treating the value of either the normal or tangential component of the flux density at each sample point as a gray level . Inverse techniques proposed to date rely largely on phenomenological models for analyzing the images to obtain estimates of the size and shape of the defect [7–10]. Unfortunately, these techniques call for considerable computing resources.