Abstract: Visual sensing of weld pool in pulsed twin arc welding of aluminum alloy is very difficult because of strong arc and reflection. A two-directional synchronous visual sensing system is established consisting of hardware and software system. By near infrared filter technology and exposure time of charge-coupled device (CCD) shutter controlling, clear weld pool images from rear and side directions are obtained. A new butterworth low-pass filtering approach is proposed to remove the horizontal stripes interference of high frequency and desired result is obtained. Through image processing methods such as gray stretch, threshold segmentation and edge detection, the profiles of weld pools from the two directions have been acquired. By analyzing the geometry of the weld pool images, it is found that the rear part of the weld pool is like parabola and the front and middle part is like ellipse. The geometric parameters of the weld pool such as the width, length, area, perimeter, rear factor and other parameters are defined and calculated on the basis of parabolic and elliptic curve equation calculation.

Key words: aluminum alloy;weld pool;twin arc pulsed metal inert-gas welding;two-directional synchronous visual sensing;butterworth low-pass filtering;geometric parameter