Research On Seam Tracking Based On Line Structured Light Vision Sensing

Welding is a traditional material joining technology,which plays an irreplaceable role in the industrial field.In recent years,with the development of industrial level,intelligent and efficient welding technology is required,so automatic welding technology has become a research hotspot.The core content of automatic welding technology includes welding seam tracking and welding control,and weld tracking is the basis of realizing efficient automatic welding.Among the numerous seam tracking sensing technologies,the line structured light vision sensing technology has the advantages of good universality,rich information,strong stability and high precision,so it is widely used in the actual welding identification and tracking.In the process of seam tracking,the diversity of weld types and interference factors(light interference,mechanical vibration,etc.)in the welding process will affect the identification of weld feature points,so the extraction and data processing of weld position information are the key problems to be solved.Therefore,the method of seam feature points recognition and seam tracking based on line structured light vision sensing was studied.Firstly,a set of weld seam tracking experiment platform based on line structured light vision sensing was established,which main contents are the design of structured light vision sensor,the design of rectifying module,the selection of control and motion module,and the introduction of weld seam tracking system software.In order to reduce the impact of lens distortion on the imaging,the imaging model was firstly introduced,and then chessboard images are used to correct the image distortion.The line structured light vision sensor was calibrated to obtain the mathematical relationship between the image pixel unit and its corresponding physic distance.Then,the feature point recognition technology of weld structured light image was studied.For V-shaped weld,a feature point recognition method based on the farthest point search algorithm was studied,which includes: image median filtering,centerline extraction based on weighted gray-scale gravity center method with weight of square stripe intensity limited by stripe boundary,and feature point recognition based on the farthest point search algorithm.For I-shaped weld and lap weld,a feature point recognition method based on improved Mean Shift algorithm was studied.After determining the initial drift point,feature points of the weld were identified directly by the data barycenter shift.Then,weld seam feature tracking based on linear kernel correlation filter tracking algorithm was studied.The tracking target in weld structured light image was determined and utilized as positive sample.Numerous negative samples were generated by circularly shift the positive sample.The positive and negative samples were used to train a classifier with kernel function by ridge regression.In a new frame of a weld structured light image,target and background were classified using the well-trained classifier,so the location of the target was obtained.Finally,for V-shaped weld,I-shaped weld and lap weld,particle filter algorithm was used to filter the weld center position,so as to reduce the influence of process and measurement errors on the seam tracking accuracy during seam tracking process.The weld center position was determined as the system variable,so the dynamic system equation was established.The colored measurement noise was virtualized as approximate gaussian white noise,and the filtering method of weld center position based on particle filter algorithm was constructed.The experimental results showed that the particle filter algorithm can significantly reduce the tracking error and improve the precision of weld seam tracking.