Fault classification based on computer vision for steel wire ropes

Abstract

Abstract As an actual steel material application, steel wire rope (SWR) is widely used in various industrial cases. However, in real cases, SWRs are inevitably damaged by erosion, friction, etc. Therefore, it is of great importance to monitor SWRs’ surface conditions in case of accidents. In recent years, computer vision recognition of SWR’s surface has been extensively studied. Based on it, we propose an image-processing method to detect local flaws of SWRs, which extracts the features from the illumination area on the surface of the SWR. Firstly, a three-feature dimension descriptor is constructed by combining the edge grayscale value gradient of the illumination area and morphological features of the illumination area. Then, the KNN (k-nearest neighbour) algorithm is used as a classification method to distinguish the local flaws. According to the experiment results, our proposed method has a high accuracy which is about 94.35%, and low time cost. It can be used in future local flaw detection.