Research On 3D Measurement Of Binocular Stereo Smooth Surface Based On Structured Light Reflection

With the continuous development of science and technology,smooth surface products,with their unique optical characteristics,have been widely used in the field of production and manufacturing,such as glass surfaces such as virtual reality(VR)lenses,smooth paint surfaces such as automobile bodies,and smooth polished metal surfaces.At the same time,the demand for three-dimensional(3D)measurement of smooth surface is generated.At present,in the field of 3D measurement,smooth surface parts are difficult to detect due to their inconspicuous surface imaging characteristics and specular reflection characteristics.The 3D measurement of smooth surface is still in the stage of contact measurement,which seriously restricts the development of parts processing and manufacturing of smooth surface.Therefore,it is necessary to study a kind of efficient,nondestructive and high-precision 3D measurement technology for smooth surface.In this paper,the high precision and high efficiency of the fringe structured light 3D measurement technology are combined with binocular stereo vision to solve the height-gradient ambiguity problem in 3D measurement of smooth surfaces.This paper has made in-depth research on system calibration,phase extraction algorithm,stereo correction and matching algorithm,and gradient integration 3D reconstruction algorithm.The main research contents include:The important research significance of 3D measurement technology for smooth surface is discussed.The advantages and disadvantages of optical 3D measurement methods such as laser triangulation method,time-of-flight method,and binocular vision method are comprehensively compared,and the necessity of developing high-precision non-destructive 3D measurement technology for smooth surfaces is proposed.The measurement principle of fringe structure light reflection is introduced,and the reason for the existence of height-gradient ambiguity in monocular measurement is analyzed.Aiming at the problem that binocular vision cannot be matched according to grayscale characteristics when applied to smooth surfaces,a 3D measurement method of smooth surfaces combining fringe structure light reflection and binocular vision is proposed.Aiming at the 3D measurement of smooth surface combining fringe structured light reflection and binocular vision,the binocular camera calibration technology was analyzed,and the phase solution and phase unwrapping algorithms of fringe structured light were studied.Considering that the binocular camera cannot directly observe the screen,an external parameter calibration method based on the mirror orthogonal constraint between the screen and the camera is proposed.Aiming at stereo matching on smooth surfaces,a matching algorithm based on stereo correction and the uniqueness constraint of the surface normal vector was proposed for the first time.Aiming at the problem of algorithm efficiency and system calibration error,an optimization scheme of matching algorithm is proposed,and a complete matching point search algorithm software is designed.Considering the problem of limited accuracy of binocular vision,a method of transforming surface normal vector data into gradient data in the matching results was designed.A method was proposed to improve the measurement accuracy of the system by using the gradient integration algorithm.The principles of gradient integration algorithms such as direct integration,Fourier transform integration and regional wave-front reconstruction are studied.The integral reconstruction algorithm implementation program is designed,and comparative simulation experiments are carried out to realize the 3D reconstruction of the simulation surface shape.A smooth surface 3D measurement experiment based on binocular stereo vision based on structured light reflection was established.The calibration of the measurement system was completed,and the convex spherical mirror was measured in three dimensions.Using the matching algorithm and gradient integral reconstruction algorithm,the 3D shape of the measured surface was restored.Surface fitting was performed on the recovered shape,which verified that the error between the fitted surface radius and the true value was 0.022 mm,and the root mean square error between the measured value and the fitted value was 0.0329 mm.