Welding seam defect detection for canisters based on computer vision

This paper proposed two novel method of real-time online detection of welding seam defects for canisters based on computer vision. In the proposed approaches, the region of welding seam is aligned after a preprocessing procedure to the acquired images. The first method is named as column gray-level accumulation inspection (CAI). In this method, an original curve is shaped by implementing the accumulating operation, followed by being exerted the mean value smoothing operation. Then a modified first difference method is used for the curve in order to segment the defects of the image of welding seam. The second method, named as frame difference testing (FDT), is also proposed for defect detecting. Finally, an information fusion approach based on D-S evidence theory that combines the CAI and FDT methods is used to reduce the rate of false alarm and improves the reliability of defect detection. Experimental results show that the proposed method can detect the welding seam defects with 90 percent accuracy and it can meet the requirement of real-time on-line continuous detection.