Surface Corrosion Detection and Classification for Steel Alloy using Image Processing and Machine Learning

Author Name(s): Sanjay Kumar Ahuja, Manoj Kumar Shukla, Kiran Kumar Ravulakollu
Author Email: SanjayAhuja@India.com

Abstract

Usually Stainless steel is notable for corrosion resistance; however, corrosion can happen in the environment with low oxygen and high salinity environment. The visual approach for corrosion detection is completely dependent on the quality of images and specialist’s knowledge and experience. Corrosion detection and analysis can be done either through chemical based approach or image processing based approach. Digital Image Processing is widely used non-destructive approach for detecting corrosion on the metal surface. It provides cost-effective, fast and reasonably accurate results provided appropriate algorithms have been used depending on the environmental conditions, corrosion type, lighting conditions etc. Several algorithms related to colour, texture, noise, clustering, segmentation, image enhancement, wavelet transformation etc. have been used in different combinations have been developed by different research groups for detecting corrosion using digital image processing techniques. This paper presents an adaptive self-learning approach for image processing based classification techniques for detecting different types of corrosion in steel. Corrosion has different colour and texture properties that can be analysed using image processing techniques. In the present paper, the research done in this field has been extensively reviewed to identify the challenges and the achievements in this area. For the present work, a combination of image segmentation approach along with machine learning approach has been used for classification of corrosion at the different concentration level of HCl and significantly reasonable results were obtained.

Introduction

Steel is an Iron alloy with other elements like Carbon, Chromium etc. in different percentage. The properties of steel e.g. hardness, ductility, and tensile strength varies with different percentage of element mass. It is widely used in construction and various other applications because of its high tensile strength, better corrosion resistance and low maintenance cost. Different types of corrosion are visually different and require different image processing algorithms.

Conclusion

The work has reported an average accuracy of 92.4% with field images of surface corrosion at 6 different level of HCl concentration. The texture features are deployed using GLCM algorithm for 19 properties. As an enhancement of the present work texture features were combined with other features like RGB colour and HSV properties of image to improve the accuracy

 

1,458 total views, no views today

Download PDF File

About the author: admin