Research on Seam Welding Recognition Based on Image Processing

Inhalt:
In response to the problem of unsatisfactory recognition and positioning of material seam welding caused by changes in illumination, an image processing-based seam welding recognition method is proposed, through the steps of image feature extraction, improved BP seam welding and non-seam welding image classification, seam welding image recognition and positioning, etc. MATLAB simulation shows that the image processing-based seam proposed in this study welding recognition method can be effectively identified in different lighting conditions to locate the seam welding of steel pipe material, and the detection rate with overlap recognition reached 91.5%, with a false alarm rate of 7.2%, and the detection rate without overlap recognition was 86.3%, with a false alarm rate of 4.3%. Compared with the traditional ostu identification, this study method has higher detection rate and better performance.