Surface Damage Detection for Steel Wire Ropes Using Deep Learning and Computer Vision Techniques

Abstract Steel wire rope (SWR) is of great importance to its many industrial applications. When SWR is damaged, it is likely to result in serious consequences. Therefore, it is important to do research in the field of SWR damage detection. Computer vision-based surface damage detection methods for SWR can operate with high detection accuracy and good adaptation for different types of SWR. Conventional machine learning methods with manual feature extraction have strong subjectivity. If the discriminant information cannot be extracted accurately, the detection accuracy decreases. To address this problem, this paper proposes an intelligent SWR damage detection method, based on a convolutional neural network, which has powerful learning ability and can automatically extract discriminant features by training surface images of the SWR. The experiment results show that the proposed method, based on deep learning, has a higher F1 score and a higher detection speed than four other conventional machine learning methods.