Research on Surface Damage Detection Method of Mine Hoist Steel Wire Rope Based on IDeblurGANv2-YOLOv7

A surface damage detection method for mine hoist steel wire ropes based on IDeblurGANv2-YOLOv7 is proposed to address the issue of blurred monitoring images caused by factors such as wire rope movement and equipment vibration during inspection, which affects the effectiveness of surface damage detection. Firstly, using the DeblurGANv2 deblurring algorithm as the baseline model, the feature pyramid network is reconstructed through a lightweight MobileNetv3 network structure to achieve lightweight improvement of the DeblurGANv2 backbone network; Secondly, the multi-scale sensing patch loss function is introduced into the model, and the weight of the joint loss function is redesigned to improve the image generation performance of the network generator and effectively enhance the restoration effect of degraded images; Then, in the inference and prediction part of the model, a fast and flexible denoising method called FFD is integrated to eliminate the “chessboard” effect of the generated image and improve the generalization and robustness of the network. Finally, damage detection is performed on the restored steel wire rope monitoring image using the YOLOv7 object detection algorithm. Through experimental verification, the method proposed in this article can restore images with richer details, fewer artifacts, and stronger authenticity in a lighter network structure. It can overcome the problem of low-quality degradation of steel wire rope monitoring images and achieve efficient damage detection of steel wire rope surface health.