Application of support vector machine to image segmentation of infrared thermal waving inspection

As a key part of the infrared thermal waving non-destructive testing technique, the thermal wave image segmentation plays an important role in the efficient detection and accurate evaluation of the structural defect. In order to minimize the influence caused by the noisy background and low contrast, the support vector machine was applied to the thermal wave image segmentation. Combining with the Wiener filter, the proposed procedure pre-processed the thermal wave image at first to enhance the contrast. Consequently, several pixel values of the background and target regions were respectively chosen to compose the characteristic vectors and input to the support vector machine, whose kernel function was set to being radial based function. Finally, the classifier obtained by the training step was applied to the thermal wave image and a binary image was obtained, which had been carried out the thermal wave image segmentation. Experimental results show that the proposed method can efficiently enhance the contrast between the background and target regions with a powerful noise retraining capability. Compared with the image segmentation method based on the hard threshold, the proposed procedure is of more benefit to the identification and evaluation of the defects and is valuable for the engineering application.