Analysis of Multilevel Otsu Thresholding for Corrosion Rate of Anodised AZ91D Magnesium Alloy ​

Abstract

Magnesium and its alloy are renowned for their high strength-to-weight ratio, excellent machinability and good castability, making it a favourable material for manufacturing various automotive parts. But when exposed to the environment, magnesium and its alloy have a weak ability to resist corrosion. AZ91D magnesium alloy was anodised and immersed in a salt solution to accelerate deterioration. The image of the corroded surface resulting from the immersion test was analysed by quantising the corroded regions. This paper studies the feasibility of an image processing approach using the multilevel Otsu thresholding method by quantising the corroded region on the AZ91D magnesium alloy. This method converts the original colour image to a greyscale image and then converts it to a binary image and RGB image for the multilevel thresholding. The pixel distributions from the multilevel Otsu thresholding display a trend that, with a higher corrosion rate, the pixel’s frequency of occurrence shifted more to the darker regions. This conforms with the Otsu thresholding, where the darker region is identified as the corroded area of the AZ91D magnesium alloy. By implementing the negative image, the binary image is segmented into non-corroded and corroded areas and labelled as 0 (black) and 1 (white,) respectively. As for multilevel Otsu thresholding, the RGB image is assigned in three colours representing different corrosion severity regions. The regions of the corrosion severity are categorised into severe corrosion, moderate corrosion and low corrosion, coloured in dark blue, light blue, and yellow, respectively. Thus, the trend of the corrosion activity can be seen by implementing the multilevel Otsu thresholding and correlating it with the corrosion rate value. Furthermore, obtaining data on corrosion attack severity can provide helpful information in predicting a component’s failure and future maintenance planning.