Automated Defect Detection of Weldments and Castings using Canny, Sobel and Gaussian filter Edge Detectors: A Comparison Study

Today non -destructive-testing (NDT) has found growing applications in lots of industrial areas. By progress of technology and improvement of processing systems, a variety of computer algorithms have been proposed specified for their application in NDT. The purpose of these algorithms is to reduce human intervention in all industrial aspects as much as possible. The radiographical image interpretation is one of the most important applications. An experienced person with highly trained eyes and mind is responsible for radiographic image interpretation; therefore, the output is highly sensitive to the human experience and would suffer from bias and mistakes. Additionally, the cost of training an experienced interpreter is very high. Thus, an automated system with high reliability of interpreting the radiographic images is required. One of the most attractive areas in radiographical image processing is automated defect detection of weldments and castings. Several algorithms on edge detection and pattern recognition have been proposed in the literature. In this paper, we do a comparison study of the performance of three well -known edge detection algorithms in the context of radiographic image processing. Sobel, Canny and Gaussian filter edge detectors were implemented and applied on a welding image database. At the end, the results are shown.

Keywords:

Non-Destructive Testing , Edge enhancement , Defect detection , Gaussian filter , Image processing