Research on Surface Defect Detection of Aluminium Based on Improved Faster RCNN

Aiming at the problems that the traditional detection algorithm does not have a high recognition rate of defects on the surface of industrial aluminium and is inaccurate in locating small defects, an improved Faster RCNN deep learning network is proposed to detect eight kinds of defects on the surface of aluminium. Firstly, the feature pyramid network (FPN) structure is added to the backbone network to enhance the feature extraction capability for small defects, then the ROI Align algorithm is used instead of the rough ROI Pooling algorithm to obtain more accurate defect localisation information, and finally, the multi-attention mechanism is integrated into the feature extraction part to improve the detection accuracy of PenLiu and CaHua, with 12% improvement for PenLiu and 14.7% improvement for CaHua. The experiments show that the mean average precision mean value (mAP50) of the improved network for the detection of surface defects in aluminium is 88.58%, which is 6.23% higher than that of the original Faster RCNN network, and the ability to detect small defects in aluminium is also significantly improved.