Vision Sensing System Of Measurement For Multi-layer And Multi-passes Welding Based On Structured Light

With the development of the manufacturing equipment,the multi-level and multi-passes welding has been a commonly used method.Welding seams often have some defects,and the quality of the welding affects the reliability of the product.The detection of the appearance of the weld seam is conducive to finding the defects inside the weld.And we can judge the rationality of the welding procedure specification through the appearance of the weld seam.The traditional weld inspection has the disadvantages of long time,tool wear,small measure range.With the improvement of industrial automation,optical measurement has received more and more attention.In this paper,we study the appearance detection technology based on the principle of triangle measure.The main research contents are as follows:(1)According to the triangulation principle of line structure,we design a visual sensor system model.The visual sensing system includes camera,lens,filter,laser,embedded processing unit,etc.(2)We calibrate the visual sensing system according to the mathematical model of the system.We adopt the Zhang's template calibration to calibrate the camera parameters and the TSAI-LENZ calibration method to calibrate the relationship between robot and camera.The equation of optical plane is obtained by fitting the data point of centerline.(3)Aiming at the requirement of accuracy and real time,we adopt method of Hessian Matrix to extract the centerline.The performance of real time of algorithms is guaranteed by the parallel computing power of GPU.In the preprocessing of image,we locate the region of the light bar accurately by calculating the image projection.The region of the light bar reduces the amount of calculation.The interference of welding spatter can be removed by filtering and the Bitwise-AND Operator between the adjacent frames.We determine the threshold by using the modified information of width.We improve the method of Hessian Matrix by introducing templates of different resolution.After extracting the centerline,we repair the broken line in the centerline.Then we use the difference of angle to identify the feature points in the weld section.We obtain the three-dimensional coordinates of the centerline through the mathematical model of the visual sensing system,and reconstruct the surface of the weld surface.(4)We use the application framework Qt,the Point Cloud Library,the visual library OpenCV,the interface library of robot and the software development kit of camera to develop a graphical interface application,which can display the data,control devices,measure the size of weld seam.Finally,we verify the consistency between the artificial measurement and visual measurement by comparison experiment and analyze the probable source of the error.