Fourier Transforms for Recognition of Weld Defects in Industrial Radiography

Weld strength is one of the most important goals in the welding industry. Non-destructive testing methods such as Radiography testing (RT) are widely used for weld strength inspection. In RT test, discontinuities and defects appear on the radiograph and are surveyed by interpreters. To increase the contrast of the radiographs, image processing methods are used to sharpen the edges and identify the defects by reducing the fogginess. In this research, two different Fourier algorithms, namely, discrete Fourier transform, and cosine Fourier transform have been used to improve the contrast of welding radiographs and to identify defects and their information. The comparison of the results of the two methods shows that the discrete Fourier transform method has a better efficiency than the cosine Fourier transform method in determining the information of the welding region, IQI lines and welding defects, and its execution time is also less. The evaluation of results by experts show that the contrast of the reconstructed images improves between 5 and 40% for different welding region with the discrete Fourier transform method. According to experts’ opinions, the reconstructed images with the cosine Fourier transform method do not have the necessary efficiency due to the artifact.