Three Dimensional Sensing Of Weld Pool Surface In GTAW Based On Structured Light Reflection

A new method is studied to observe the three-dimensional weld pool surface using structured light. It's based on the principle of reflection law. The low power laser projects parallel structured light onto the weld pool surface. And then the reflection of the laser is intercepted by an imaging plane. The reflected images are captured by a camera. The reflected laser stripes change with the fluctuation of weld pool surface. The arc is suppressed using two techniques: the arc radiation decays very fast with the travel distance while the reflection of the projected laser almost remains its intensity as the laser is coherent and directional, therefore, the arc can be suppressed effectively if the imaging plane is placed at a rational distance from the arc; the band-pass filter technique is also used to suppress the arc.Based on the weld pool observation method, a 3D structured light sensing system has been designed. The sensing system is made up of a 24 mW illumination laser at wavelength of 685 nm with 9 lines, a band pass filter of 10 nm bandwidth centered at wavelength of 685nm, a camera, a frame grabber DH-VRT-CG210, a computer. The sensing system is simple and can be easily realized with low cost. Experiments are carried out to sensing the weld pool surface and the clear reflection images are captured.According to the characteristic of sensing system and the parameters that should be calibrated, the sensing system is divided into two parts: camera calibration and structured light calibration. The intrinsic and extrinsic parameters of camera are obtained using the Zhang Zhengyou's camera calibration method. A new algorithm of structured light calibration is proposed and the structure parameters of sensing system are obtained using the results of camera calibration and other measurement means.The image processing programs for the reflected images are developed, including the image smoothness method, the threshold method, the fringe pick-up method, the stripe skeletonizing method. Using the methods above, the image noises are eliminated and the feature information for the 3D weld pool surface recovery is obtained. A new algorithm is proposed for weld pool surface three-dimensional recovery. Based on the characteristic of weld pool, the corresponding relationship among the laser stripes is obtained. The corresponding relationship between the intercepted points on the imaging plane and the projected points on the weld pool surface is obtained by dispersing the image laser stripes and the projected laser stripes. According to reflection law of light, the depth of the weld pool is adjusted by using the slope as iterative variable. Finally, an engineering solution of the weld pool surface is obtained.