Research On Seam Image Processing For CO₂ Horizontal Position Welding Based On Laser Vision Sensing

With constant expanding of the quantity and scale of tall steel building construction project, how to improve the quality and efficiency of steel structure welding and reduce the labor intensity of welder has become a key issue to be resolved. To some extent, the problem can be solved by choosing the appropriate welding methods, welding materials and welding power source, etc. But the most important solution is to achieve mechanization and automation of construction steel structure welding.Seam tracking is an important element to realize welding automation, especially automatic welding process control. For the practical engineering need, the research target of this paper is to realize automatic laser vision seam tracking of horizontal position welding of large and thick plate. According to the spatter in CO2 arc welding process and other special circumstances, vision sensor design, seam image processing methods and deviation control methods of laser vision seam tracking system are mainly studied.First, hardware composition and soft flow of seam image acquisition and processing system are determined based on the extensive reference to domestic and international research about laser vision seam tracking system. The slanted laser device and vertical CCD camera are fixed in the vision sensor which is the core component of system. The angle of laser light and CCD camera is about 20. And the models of coordinate transformation between workpiece and camera with arbitrary angle are presented.Second, the seam image processing is emphatically researched. The adaptive median filter is used to remove most of the seam image noise. Wavelet coefficients of different scale and frequency band can be obtained by using wavelet transform of filtered seam image. Then, through the respective enhancement of these wavelet coefficients, contrast enhancement of seam image can be realized while suppressing the noise. The threshold of seam image binarization can be obtained through curve fitting of the average gray level within laser region and the result of Otsu algorithm based on least square method, thereby realizing the adaptive threshold selection according to different seam images. Third, the algorithm of anti-noise-dilation-erosion morphological edge detection is introduced to realize edge detection of preprocessed seam image. Using a circular structure element with radius 5, the edge of seam image can be detected while the spatter and noise are effectively eliminated. Then, the centerline of laser region is effectively extracted by using the morphological skeleton and average contour with the removal of isolated noise. Subsequently, the feature points of groove and seam, which are used to calculate deviations, can be rapidly detected using slope analysis method.Fourth, research on deviation adjustment is carried out using the dual-mode control strategy of adaptive fuzzy PID and fuzzy algorithms. With the research mentioned above, real-time seam tracking and torch position adjustment can be achieved.Last, the seam tracking results of actual single-layer and single-pass and multi-layer and multi-pass CO2 horizontal position welding testify the effectiveness and reliability of proposed seam image processing and control methods.