Defect detection algorithm of wire rope based on color segmentation and Faster RCNN
Wei Li, Tianxin Dong, Haibin Shi, Lei Ye
Abstract
Aiming at the four common types of wire rope defects in tower crane operation, such as deformation, core extrusion, steel wire extrusion and surface wire breakage, a wire rope defect detection algorithm based on color segmentation and faster region convolution neural networks (Faster RCNN) is proposed. The color segmentation algorithm is used to extract the wire rope, if there is a big difference from the normal shape of the wire rope, it is directly judged as deformation defect; if there is no difference or small difference, the Faster RCNN network is used for detecting the defect in detail. Experiments show that the average detection accuracy of the algorithm reaches 90.61 %, the defective parts of wire rope can be detected effectively, thereby ensuring the safe operation of the tower crane.