ABSTRACT Eddy current pulsed thermography (ECPT) technique has been investigated for steel corrosion detection and evaluation. However, the random interference was not considered and disposed adequately, which bedims the original thermal images and reconstructed feature ones. This paper applied bidimensional empirical mode decomposition (BEMD) method which used to denoise the thermal image sequences, and the contrast between corroded and uncorroded regions in the skewness reconstructed images of 6-months and 12-months samples was improved by 19% and 28%, respectively. By this means, the skewness of the temperature response curve of each pixel was extracted to reconstruct the feature image. Furthermore, the differential skewness and the contrast of reconstructed skewness image are utilised for distinguishing the early corrosion and the moderate corrosion. The results demonstrate that the method provides higher sensitivity and accuracy in corrosion detection with ECPT.
Read full abstract