Three-dimensional Reconstruction Of Weld Seam And Weld Spatter Onsite Detection Based On Machine Vision

During the operation of automatic welding line,due to the blockage of the nozzle of the welding gun,it is very easy to produce the surface defects of the welding products,which will seriously affect the use of the finished product and the operation of the welding line.Artificial detection method is a common detection method in automatic welding line.It has the disadvantages of low detection efficiency and precision and great influence on the physical condition of the inspectors.It is unable to meet the requirements of high speed welding automation production line.The non-contact detection method based on machine vision has developed rapidly.It has the advantages of fast detection speed,high precision and good stability.It can make up the shortcomings of artificial detection.Based on the method of machine vision,this paper uses three-dimensional imaging technology to study the three-dimensional reconstruction process of the feature points of the smooth surface of the non chromatic surface,designs the three-dimensional reconstruction algorithm of the weld,studies the extraction process of the signal in the frequency domain by the image transformation technology,designs the position detection method of the welding spatter,and verifies the algorithm by the experiment.Correctness provides a strong guarantee for improving the detection level of welding automatic production line.A Three-dimensional reconstruction algorithm based on machine vision is designed.In view of the difficulty of extracting the features of the smooth surface without chromatic aberration,a surface feature supplement method is designed.A binocular vision system model is established to improve the capture efficiency of the image.A set of preprocessing algorithm is designed to solve the three-dimensional reconstruction problem of the smooth surface without chromatic aberration.The extraction algorithm of feature points is used to improve the precision of feature point extraction.According to the matching problem of the smooth surface,the matching algorithm of feature points is designed according to the position relation of the feature points and the angle of grid distortion,and the exact matching feature points are obtained.The experimental results show that the feature point extraction and matching algorithm of non chromatic smooth surface has good robustness,high accuracy,high accuracy and repetition rate of matching feature points,the distance error between the feature points after three-dimensional reconstruction is less than 0.05mm,and the standard deviation is 0.0362mm.A complementary method of matching feature points to the density is designed,and the three-dimensional coordinates of the feature points are obtained.The three-dimensional reconstruction of the feature points of the smooth surface without chromatic aberration is realized by combining the camera calibration information and the least square method.The three-dimensional reconstruction algorithm of the feature point of the smooth surface based on the machine vision is applied to the three-dimensional reconstruction of the feature points of the weld surface.The distribution of the feature point cloud conforms to the actual fish scale distribution on the weld surface.The accuracy of the reconstruction meets the precision requirements of the high quality welder at the present stage,and can be applied to the defect detection of the weld surface.A position detection algorithm based on machine vision is designed to adhere to the spatter on the surface of welded plate.In view of the problem of obtaining the weld plate image,a monocular vision system model is set up.In view of the problem of image information conversion,the preprocessing method of two-dimensional adaptive Wiener filtering is selected to smooth the noise area and sharpen the spatter area,and the Fourier transform is used for the problem of the spatter segmentation of the weld plate.The method converts the image to the frequency domain,designs an elliptical window high pass filter according to the amplitude distribution law of the spectrum,divides the high frequency spatter information,ensures the segmentation precision of the splash,and designs the weld area filtering algorithm based on the region of interest in the image,and uses the threshold method to solve the problem of the splash.The spatter information that adhered to the weld plate was extracted.The experimental results show that the onsite detection algorithm based on the machine vision adhered to the surface of the weld plate is 0.9,the extraction accuracy is 0.9669,and the advantage is obvious compared with the traditional defect detection algorithm.It can be applied to the detection of the spatter on the weld plate surface.