Research On Three-dimensional Visual Inspection Method Of Laser Weld Based On Mesh Structure

Welding plays a very important role in the rapid development of industrial intelligent manufacturing in China,so industrial production also has higher requirements for the quality of welding.At present,post weld seam detection mainly relies on manual visual observation and simple measurement of detection tools.It belongs to contact measurement with obvious disadvantages and does not meet the requirements of current industrial production.Due to many problems in contact measurement,non-contact visual inspection system will gradually replace manual inspection and become the mainstream weld inspection technology in the future.At present,many effective methods have been developed in non-contact visual appearance detection at home and abroad.Among them,structured light three-dimensional vision method is a three-dimensional visual detection technology based on triangulation.The laser used has the advantages of high brightness and good coherence.The surface information of the object to be measured is very rich,and the construction of the system is relatively simple.However,when using single line laser as auxiliary structured light,the detection efficiency is low.Therefore,the three-dimensional detection technology using multi line and grid laser as auxiliary structured light needs to be further explored.By analyzing the current weld visual appearance inspection systems at home and abroad,this paper uses the structured light three-dimensional visual inspection method to detect the weld,uses this method to obtain the height of the weld feature points,and then reconstruct the weld surface.In order to obtain the information of the weld surface,the hardware and software of the weld appearance inspection system are developed based on the principle of laser triangulation,and the calibration algorithm is used to calibrate the system.When the grid laser light source shines on the weld,the grid laser stripe at the weld will be deformed.At this time,the weighted average method,Gaussian filter The maximum entropy method,skeletonization and self-developed burr removal algorithms in the automatic threshold method process the collected weld image to obtain the grid centerline with single pixel width,and then use the detection system to identify the weld feature points to complete the three-dimensional reconstruction of the weld.However,when the angle between the grid laser and the camera is small,the deformation of the laser line is small,and the weld feature points are difficult to extract.Therefore,in view of the above situation,this paper changes the field of view of the camera and the angle between the grid laser and the camera by adding a reflector,so as to obtain more accurate height information of feature points on the weld surface,and then fit the height information and the extracted coordinate information with the B-spline curve method,and finally obtain the three-dimensional contour of the weld.