Rust Defect Detection and Segmentation Method for Tower Crane

In view of the long-term exposure of tower crane to the natural environment, which is likely to cause rust and lead to the problem of construction safety, this paper proposed a rust defect detection and segmentation method for tower crane.Firstly, the image preprocessing is used for denoising, normalization and histogram equalization to enhance the overall contrast of the image. Secondly, improve the YOLO V3 algorithm and introduce SENet, the channel attention mechanism, to make the rust feature information more prominent. Finally, use threshold segmentation to segment and extract the rusted area from the improved YOLO V3 recognition result to obtain the final rusted area. The experimental results show that the improved YOLO V3 algorithm mAP improves by 2.23% and improves the detection accuracy. This method can effectively detect and segment the rusty area of the tower crane.