Research On 3-D Vision Measurement Technology Of Wheel Hub Based On Monocular Line Structured Light

With the rapid development of automobile parts manufacturing industry,in the field of industrial manufacturing quality inspection,more and more attention is paid to the research of three-dimensional measurement technology for the full size,surface topography and defects of automobile parts.The wheel hub is an important part of automobile parts,and the threedimensional measurement of wheel hub is of great significance for avoiding the failure of wheel hub and ensuring the safety of automobile travel.The traditional contact-type 3D measurement technology cannot meet the requirements of high-precision,high-efficiency,and non-destructive online inspection.The on-line measurement technology of non-contact parts based on machine vision is becoming a research hotspot of quality inspection in automobile parts production line.According to the characteristics of hub parts,this paper uses line structured light projection and monocular vision to study the measurement of outer contour morphology and upper edge location of weld on hub surface.The research work of this paper mainly includes the following aspects:Firstly,based on the analysis of the principle of three-dimensional vision measurement with linear structured light,the corresponding three-dimensional vision measurement scheme is proposed for the reconstruction of the outer contour of the hub and the positioning of the upper edge of the hub weld;Through the computer vision technology based on structured light projection,a monocular rotating vision system based on line structured light scanning is designed;A monocular rotation calibration method is proposed to solve the problem of monocular camera calibration of hub axisymmetric contour;Aiming at the problem that part of the line structured light projection fringe on the wheel hub is missing,the center line extraction algorithm of light stripe based on ROI gray adaptive enhancement is proposed to realize the line projection image extraction of wheel hub outer contour with high robustness.It provides a technical scheme for the rapid measurement of the outer contour of automobile hub.Secondly,aiming at the problems of large volume and inconvenient adjustment of the traditional visual measurement system,through the embedded system development technology,an embedded wheel hub three-dimensional visual measurement system based on Zynq is designed to realize the control and data processing functions of embedded threedimensional visual measurement from the hardware design of PL part and the software design of PS part,Provide embedded processing solution for on-line rapid measurement of wheel hub.At the same time,the monocular structured light 3-D measurement system designed in this paper is used to carry out the surface 3-D reconstruction experiment on the standard metal gauge block and the standard bat.The reconstruction error of the gauge block plane,the two ball radius of the standard bat and the error of the ball spacing are calculated to evaluate the accuracy of the measurement system.The measurement accuracy of the metal gauge block surface is better than 0.05 mm and the bat measurement accuracy is 0.06 mm.Through the above experiments,the research work of this paper is verified.Finally,aiming at the problem of positioning and measuring the upper edge of hub weld,an improved dual input U-Net semantic segmentation model is designed through deep learning technology,which can effectively segment the weld area of hub weld image.Compared with the traditional canny segmentation and U-Net model,the MIo U is improved by 2.9807% and 0.7152% respectively.The effectiveness of the model is verified by the positioning measurement experiment of the upper edge of the hub weld.Through the research of wheel hub 3-D vision measurement technology based on monocular structured light,the problems of outer contour morphology measurement and upper edge positioning measurement of weld on wheel hub surface can be effectively solved.