Three-dimensional Visual Sensing Of Gmaw Pool Using Grid Structured Light

With the rapid development of computer technology, automation technology and and artificial intelligence, welding technology is required to be more automatic and more intelligent.The quality control during the welding operation is the core issue of weld intelligentizing.To achieve that purpose,it’s essential for detecting the effective information that can reflect the welding quality. The three-dimensional information of the weld pool surface relates to the weld appearance, the mechanical properties of the welded joints and the back penetration of the weld directly.For the speclular feature of the weld pool surface,this paper adopts the active visual sensing technology to detect the weld pool of GMAW and reconstruct the weld pool surface using reflected image form grid structured light.Firstly, a three-dimensional visual sensing system has been established. The system includes three components:motion control part, visual sensing part and welding operation part. The system software is programmed to realize real-time image acquisition and welding control.Combining the characteristics of the structured light measurement, the effect of laser beam parameters, structure light type is analyzed, and the selection principle is determined.The weld pool surface model, grid structured light model, weld pool surface reflection model and reflected image model is constructed.Based the models,the simulation of structured light active visual sensing process is built.In the simulation system, structured light three-dimensional measurement is conducted to static weld pool surface.According to the simulationn results,the influence of the weld pool geometry and sensing system parameters to the reflected grid pattern is analyzed.The recognition and processing program of reflected image is developed. In the image recognition module, the average gray level of the image and the image signal characteristics in frequency domain are combined to recongise the reflected image with grid pattern. Low pass filter, top-hat transform, gray stretch are adopted for removing noise and enhancing the quality of the reflected image. The local adaptive thresholding method is adopted to binarizing the image. The skeleton structure of the grid pattern from the reflected image is extracted.Camera calibration is completed and the transform from pixel coordinates to world coordinate is realized.Weld pool reconstruction algorithm based on the normal vector is proposed: calculate the normal vector for the reflection point clouds data, reconstruct the surface by biharmonic spline interpolation, iterate steps above until approximating the accurate solution. Reconstruct the weld surface of the reflected data from the simulation system to analyze the error of the reconstruction algorithm. Finally, the three-dimensional weld pool surface using the reflected image captured in the actual welding process is carried out. The feasibility of the active visual sensing technology with grid structured light in GMAW is verified.