Remanufacturing of high-value engineering structures is set to become an important aspect of the future manufacturing industry. However, this depends on the ability to accurately, and rapidly inspect used components for damage, such as corrosion. Visual inspection in both manufacturing and remanufacturing is often performed manually, which is a time-consuming, subjective process. This paper looks at the application of machine learning to the automation of visual inspection for remanufacturing. A Gaussian mixture model is trained on a novel set of image features, specifically designed for the task of corrosion detection in used parts. The probabilistic model is used to segment images of automotive engine components into corroded and non-corroded areas. It is possible that the uncertainty in this segmentation may be used to automate further inspection.