Joint Research Conference
June 24-26, 2014
Recent development of image sensing technologies has facilitated process monitoring and fault detection. In various manufacturing applications such as steel, composites, and textile production processes, defect detection in noisy images is of special importance. There exist several methods in the literature for image denoising and defect detection. However, most of these methods perform denosing and detection separately, which affects the detection accuracy. Additionally, the low computational speed of these methods is an issue in real-time process monitoring. In this research, we develop a novel methodology for defect detection in noisy images that have smooth background. The proposed method decomposes an image into three components, namely, smooth background, sparse defects and noises, through a regularized least square optimization model. We also propose a fast algorithm for solving the optimization model. Using simulation, we compare the proposed methodology with exiting defect detection algorithms in terms of the accuracy as well as computational speed. Furthermore, we use photoleastic image data to validate our methodology.