Automated Grain Size Analysis in IF Steels Using Digital Image Processing

Grain size is one of the most relevant microstructural parameters governing the mechanical performance of IF (Interstitial Free) steels, as it directly influences properties such as strength, hardness, and ductility, according to widely documented relationships in the metallurgical literature [1,2]. In forming applications, small variations in grain size can lead to significant changes in the materials behavior. Traditionally, grain size determination is performed through manual metallographic methods based on established standards; however, these procedures can be subjective and dependent on the operator’s experience [3]. This work presents an automated method for determining grain size in IF steels based on digital image processing techniques. The analysis was performed using optical micrographs obtained through conventional metallographic preparation. The images were converted to grayscale, filtered for noise reduction, and segmented using edge detection, following approaches widely used in computer vision applied to materials science [4,5]. Grain size was calculated using the linear intercept method in accordance with ASTM E112, which ensures direct comparability with traditional methods [6]. The implemented algorithm begins with the normalization of image intensity to reduce illumination variations. A spatial filter is then applied to attenuate noise while preserving grain boundaries. Segmentation is carried out through edge detection, enabling the identification of dominant microstructural contours. Based on this segmentation, a grid of randomly oriented test lines is superimposed, and the number of intersections with grain boundaries is counted. Finally, the average grain size is calculated from the number of intersections per unit length, following the guidelines established in ASTM E112 [6]. The validation of the automated method was carried out by comparing it with manual measurements performed by a metallography expert on the same set of micrographs. The results showed an average agreement of 97.2%, with relative deviations below 5% in 85% of the images analyzed. The mean absolute error remained below 0.4 ASTM units. Likewise, a linear correlation coefficient of RÂ2 = 0.96 was obtained between the automatically determined values and the reference values, confirming the strong relationship between both methods and their statistical consistency [7]. During the analysis, it was observed that the quality of the metallographic etching and the lighting conditions influence the accuracy of grain boundary segmentation, as previously reported in automated microstructural analysis studies [5,8]. However, the use of adaptive thresholding techniques helped minimize these effects, improving the robustness of the method against moderate variations in image quality. From a materials analysis standpoint, this characteristic is particularly relevant, as it allows the grain size values obtained to be used as a reliable parameter in studies correlating microstructure and mechanical properties [1,9]. Overall, the validation results confirm that the proposed automated method allows the grain size of IF steels to be determined with accuracy and repeatability comparable to those of manual methods, while significantly reducing analysis subjectivity and evaluation time. This approach constitutes an effective tool for microstructural analysis in academic and research environments. Flowchart of the proposed algorithm Example of image processing and grain size calculation Comparison between the manual and automated methods