An Objective Metallographic Analysis Approach Based on Advanced Image Processing Techniques

Metallographic analyses of nodular iron casting methods are based on visual comparisons according to measuring standards. Specifically, the microstructure is analyzed in a subjective manner by comparing extracted image from microscope pre-defined templates. The achieved classifications can be confused, due fact that features human being could interpreted differently depending many variables, such as conditions observer. In particular, this kind problem represents an uncertainty when classifying metallic properties, which influence integrity castings play critical roles safety devices or structures. Although there existing solutions working with images and applying some computer vision techniques manage measurements microstructure, those results not too accurate. fact, they able characterize all specific and, cannot adapted several characterization regulation customer. Hence, order solve problem, we propose framework improve automatize evaluations combining classical machine for feature extraction deep learning technologies, objectively make classifications. To adapt real analysis environments, included inputs our models were gathered directly historical repository metallurgy Azterlan Research Centre (labeled using expert knowledge engineers). proposed approach concludes these (a classification under pipeline neural networks quality ANN classifier) viable carry out metallographic great accuracy time, it possible deploy software work real-time situations. Moreover, method provides direct way classify metallurgical molten metal, allowing us determine behaviors final produced parts.