Abstract
This paper sheds new light on the classification of material corrosion levels in an attempt to reduce the frequency of subjective judgment on it. Here, grounding grid material is taken as the study object, and the images of steel plates are collected. Based on the deep learning, the artificial bee colony algorithm is integrated to determine the best split point to carry out the corrosion area segmentation. In this way, the effectiveness of the deep learning algorithm can be improved. A material corrosion classification model is thereby built based on the SOM. The findings show that the corrosion levels can be divided into three classes, i.e. the blue part is the most severe corrosion area, the green part is the moderate corrosion area, and the red part is the least corrosion area. From the above test results, the staff can get across to what degree the relative corrosion of the grounding grid materials in the affected area reaches, then if necessary, carry out an emergency repair measures against the most severely corroded areas.
