Recognition of corrosion area using image processing technique

Igarashi, Takahiro  ; Otani, Kyohei   ; Aoyama, Takahito   

Corrosion areas were predicted using the contour extraction methods (Global, Adaptive mean (AM), Adaptive Gaussian (AG), Canny) and feature point extraction methods (KAZE, AKAZE, BRISK, ORB, AFD, FAST, MSER, SBD) on surface images of carbon steel specimens that had been subjected to repeated dry and wet tests before rust removal, and it was confirmed that both methods could approximately extract relatively heavily corroded areas. In particular, it was confirmed that the feature point extraction method can extract intense corrosion areas.