Research On 3D Profile Measurement Of Smooth Surface And 3D Point Cloud Stitching Technology

With the rapid development of modern industrial technology,smooth surface products are becoming more and more diversified,and they are more and more widely used in many manufacturing industries,improving the quality and performance of related products,and the demand for three-dimensional(3D)profile measurement of smooth surfaces is also increasing rapidly.At present,the 3D profile measurement of smooth surfaces is still in the contact measurement stage.Although this technology has high measurement accuracy,it has low efficiency and cannot realize online detection.Therefore,it is of great significance to study a convenient,fast and high-precision non-contact 3D profile measurement technology for smooth surfaces.In this thesis,the 3D profile measurement technology of smooth surfaces based on fringe reflection and binocular vision is studied,and research on system phase error and multi-view 3D point cloud splicing technology are carried out.The main research contents of this thesis are as follows:(1)The key technical links in the 3D profile measurement technology of smooth surfaces based on fringe reflection and binocular vision are studied.Firstly,the measurement principle and design of the system are introduced.Secondly,the camera calibration technology and screen calibration technology are studied.Then,the phase extraction algorithm of the fringe light is elaborated in detail,including the wrapping phase solution and the unwrapping phase.Lastly,Stereo matching algorithm based on the epipolar constraint and the uniqueness of the surface normal vector is introduced,the gradient integration algorithm is used to complete the 3D reconstruction of the surface of the object to be measured.(2)Aiming at the system phase error,it is analyzed and corrected in this thesis,which improves the measurement accuracy of the system.Firstly,gamma distortion,the main cause of phase error,is introduced in detail,and its influence on phase measurement is analyzed.After that,the gamma correction method based on Hilbert transform and the gamma active correction method based on gray-scale image projection are used to correct gamma distortion.The basic principles and specific implementation steps of above two gamma distortion correction algorithms are explained in detail,and the corresponding simulation verification of these two algorithms is carried out.(3)For smooth surface objects that cannot be measured in a single 3D measurement,a multi-view 3D point cloud stitching technology is proposed in this thesis,which realizes the measurement of the 3D profile of their surfaces.This technology does the extraction of the marker points by using the marker point extraction algorithm based on line scanning firstly,then it completes the marker point matching according to the invariance of the spatial point geometric characteristics,finally,the coordinate transform matrix Rw and Tw of the adjacent view coordinate systems will be get by quaternion method,and the multi-view 3D point cloud stitching is realized.(4)A single-view 3D contour measurement system for smooth surfaces and a multi-view 3D point cloud stitching experiment system were established,camera calibration and screen calibration were completed,and corresponding verification experiments of above theories were completed:(a)the experimental verification of the two gamma distortion correction algorithms was carried out,the phase error after correction was obtained respectively.The comparison of the phase error with the uncorrected phase error proved that the correction algorithm can improve the measurement accuracy;(b)two 3D measurements were carried out for the standard plane mirror at two locations,of which the distance is 10mm.The measurement result is 10.093mm,which is 0.093mm apart from the actual distance;(c)the multi-view 3D measurements were carried out for the convex spherical mirror,and the multi-view 3D point cloud stitching technology was used to complete the three-dimensional measurement of the convex mirror.Then,a spherical fitting was performed on the measurement data,and the fitted spherical radius of the convex mirror is 122.316 mm.Similarly,a spherical fitting was performed on the data measured by CMM,and the fitting spherical radius is 122.425mm,the difference between the two is 0.109mm.