-
测量系统由工业相机、投影仪、漫反射光屏、被测镜面、计算机组成,如图 1所示。投影仪将由计算机生成的亮度呈正弦变化的条纹投影至漫反射光屏上,光屏上的条纹经过平面镜反射成像被相机获得。其中相机获得的条纹保持最大灰度值并且不过饱和,BABAIE等人[10]提出采用整体匹配关系产生光栅; WADDINGTON等人[11]采取减少投影强度来解决图像饱和问题。本文中采用自适应条纹投影[12],通过求解相机也投影仪响应曲线来求取最佳投影灰度值。
T为光屏上任一点,设入射光线和反射光线的单位方向向量分别为i和r,P为T点经镜面反射后在相机的成像点,可以求得被测镜面反射点M的实际法线向量n:
$ \mathit{\boldsymbol{n}} = \frac{{\mathit{\boldsymbol{r}} - \mathit{\boldsymbol{i}}}}{{\left| {\mathit{\boldsymbol{r}} - \mathit{\boldsymbol{i}}} \right|}}{\rm{ }} $
(1) (1) 式为点的实际法线向量,与该点法向量的理论值比较,得到镜面在该点的面形误差。理论法线向量可由被测镜面面形方程确定[13]。
-
P点坐标在相机坐标系下可直接求出。在相机针孔模型中,P, O1, M这3点共线,在已知相机外参和被测镜面形方程,当P坐标和相机光心O1的坐标确定后,M坐标由线PO1与被测镜面的面形方程联立得到。
T点坐标可由条纹相位得到,常见的条纹相位解法有最小二乘相位估计[14]、加窗傅里叶分析法[15-16]、四步相移法[17-18],本文中采用四步相移法计算条纹相位。在图 1中,光屏上的正弦条纹由电脑通过投影仪投射至光屏,投射的横、竖正弦条纹如图 2a、图 2b所示。
图 2a中光屏任意一点的光强分布为:
$ I = {A_0}{\rm{cos}}(2{\rm{ \mathsf{ π} }}x/p)\; $
(2) 式中, A0为光强幅值,p条纹间距,x为横向坐标值。光屏上的条纹经过被测镜反射后,在相机的像素平面上任意一点P的光强表示为:
$ \;I = {I_0} + {A_0}{\rm{cos}}\phi $
(3) 式中, I0是系统引入的背景噪声,像平面上的P点对应光屏上的T点的相位相等。首先求出P点的相位值Φ,再据相位相等,求出T点的坐标值。为了求出Φ,需要应用相移算法,加上相移算子δ后,(3)式可展开为:
$ I = {I_0} + {A_0}{\rm{cos}}\phi {\rm{cos}}\mathit{\delta } - {A_0}{\rm{sin}}\phi sin\mathit{\delta } $
(4) 采用四步相移算法,即令δ分别取值为0,π/2,π,3π/2,(4)式展开化简为:
$ \left\{ \begin{array}{l} {I_1} = {I_0} + {A_0}{\rm{cos}}\phi \\ {I_2} = {I_0} - {A_0}{\rm{sin}}\phi \\ {I_3} = {I_0} - {A_0}{\rm{cos}}\phi \\ {I_2} = {I_0} + {A_0}{\rm{sin}}\phi \end{array} \right. $
(5) 可得到T点的相位值:
$ \phi {\rm{ = arctan}}\left( {\frac{{{I_4} - {I_2}}}{{{I_1} - {I_3}}}} \right)\; $
(6) 在光屏坐标系中T点的x坐标值为:
$ x = \phi \times \frac{{2{\rm{ \mathsf{ π} }}}}{p} $
(7) 同理,将投射到光屏上的竖条纹改为横条纹如图 2b所示。重复上述计算过程可得到T点的纵向坐标值y,可以得到像素平面上任意P点相应的光屏上T点的位置坐标。由(6)式所获得的相位是由反正切函数值表示,根据反正切函数的固有性质,这些相位值仅处于[-π/2, π/2]区间内,把不连续的相位采用相位模2π扩展获得准确的Φ值,才能得到准确的x, y坐标值。
-
求出的P, M, T坐标分别位于相机坐标系、世界坐标系、光屏坐标系下,现将P, T的坐标统一到世界坐标系下。如图 1所示, 世界坐标系为XYZ,光屏坐标系为XsYsZs,相机坐标系为XcYcZc。
设光屏坐标系与世界坐标系的变化矩阵为N1, 相机坐标系与到世界坐标系的变化矩阵为N2。在被测镜面上贴上若干个不在一条直线上的标志点,获得这些标志点在相机坐标系和世界坐标系下的坐标,采用单片空间后方交会算法[19]可计算得到N2。N1的求解过程为分别求解旋转矩阵R、平移矩阵F。A,B,C为光屏坐标系XsYsZs坐标轴上的3点,其中A为原点,B为Xs轴上的一点,C为Ys轴上的一点,设在世界坐标系中它们的坐标分别为:A(x1, y1, z1),B(x2, y2, z2), C(x3, y3, z3)得到光屏坐标系的Xs, Ys轴在世界坐标系中的单位方向向量$ %\overrightarrow {\mathit{AB}} $, $ %\overrightarrow {\mathit{AC}} $。
设光屏坐标系变换到世界坐标系的旋转矩阵为:
$ \mathit{\boldsymbol{R}} = \left[ \begin{array}{l} {a_1}\;\;\;{a_2}\;\;\;{a_3}\\ {b_1}\;\;\;{b_2}\;\;\;{b_3}\\ {c_1}\;\;\;{c_2}\;\;\;{c_3}\; \end{array} \right]{\rm{ }} $
(8) 则有:
$ \left\{ \begin{array}{l} \left[ {{a_1}\;\;{a_2}\;\;{a_3}} \right] = \overrightarrow {AB} \\ \left[ {{b_1}\;\;{b_2}\;\;{b_3}} \right] = \overrightarrow {AC} \end{array} \right. $
(9) (c1, c2, c3)为光屏坐标系的Zs轴在世界坐标系中的单位方向向量,由坐标轴的正交性和单位向量的性质有:
$ \left\{ \begin{array}{l} {a_1}{c_1} + {a_2}{c_2} + {a_3}{c_3} = 0\\ {b_1}{c_1} + {b_2}{c_2} + {b_3}{c_3} = 0\\ {c_1}^2 + {c_2}^2 + {c_3}^2 = 0\; \end{array} \right. $
(10) 解(10)式求得c1, c2, c3, 这样就得到了光屏坐标系变换到世界坐标系的旋转矩阵R,平移矩阵F为:
$ \mathit{\boldsymbol{F}} = \left[ \begin{array}{l} {t_1}\\ {t_2}\\ {t_3} \end{array} \right]\; = \left[ \begin{array}{l} 0\\ 0\\ 0 \end{array} \right] - \mathit{\boldsymbol{R}}{\rm{ }}\left[ \begin{array}{l} {x_1}\\ {y_1}\\ {z_1} \end{array} \right]\;\; $
(11) -
对被测镜面的面形评价参量是法线偏差[20],即为被测点的测量法线与镜面面形方程在该点的理想法线的夹角偏差。测量点的实际法线表达如(1)式所示,理想法线由被测镜面方程给出。根据求解的N1, N2将P点和相机光心O1以及T点坐标统一变换到世界坐标系中,坐标分别表示为O1′(a′, b′, c′), P′(x4′, y4′, z4′), T′(x5′, y5′, z5′)。直线P′O1′的方程为:
$ \frac{{x - a\prime }}{{{x_4}\prime - a\prime }} = \frac{{y - b\prime }}{{{y_4}\prime - b\prime }} = \frac{{z - c\prime }}{{{z_4}\prime - c\prime }}{\rm{ }} $
(12) 联立直线P′O1′方程和被测镜面方程,即可求出在M点在世界坐标系中的坐标M′(x6′, y6′, z6′)、入射光线T′M′的单位方向向量i、反射光线M′P′的单位方向向量r。则点M′的法向单位向量在X, Y, Z方向的分量表示为:
$ \left\{ \begin{array}{l} {n_x} = \frac{{{r_x} - {i_x}}}{{\sqrt {} {{({r_x} - {i_x})}^2} + {{({r_y} - {i_y})}^2} + {{({r_z} - {i_z})}^2}}}\\ {n_y} = \frac{{{r_y} - {i_y}}}{{\sqrt {{{({r_x} - {i_x})}^2} + {{({r_y} - {i_y})}^2} + {{({r_z} - {i_z})}^2}} }}\\ {n_z} = \frac{{{r_z} - {i_z}}}{{\sqrt {{{({r_x} - {i_x})}^2} + {{({r_y} - {i_y})}^2} + {{({r_z} - {i_z})}^2}} }} \end{array} \right. $
(13) 如果某测量点的测量法线方向为(nX, nY, nZ),由被测镜面方程所求得的理想法线方向为(nX′, nY′, nZ′),则该点沿个方向的法线偏差分别为:
$ \left\{ \begin{array}{l} {d_x} = \frac{{{n_X}}}{{\sqrt {{n_X}^2 + {n_Y}^2 + {n_Z}^2} }} - \frac{{{n_X}\prime }}{{\sqrt {{n_X}{\prime ^2} + {n_Y}{\prime ^2} + {n_Z}{\prime ^2}} }}\\ {d_y} = \frac{{{n_y}}}{{\sqrt {{n_X}^2 + {n_Y}^2 + {n_Z}^2} }} - \frac{{{n_Y}\prime }}{{{n_X}{\prime ^2} + {n_Y}{\prime ^2} + {n_Z}{\prime ^2}}}\\ {d_z} = \frac{{{n_Z}}}{{\sqrt {{n_X}^2 + {n_Y}^2 + {n_Z}^2} }} - \frac{{{n_Z}\prime }}{{\sqrt {{n_X}{\prime ^2} + {n_Y}{\prime ^2} + {n_Z}{\prime ^2}} }}\; \end{array} \right. $
(14) 由(14)式就能求出M点在世界坐标系中沿X, Y, Z方向的法线偏差。同理,可求出被测平面任一点的法向偏差,从而可以对被测平面进行评价。
-
为验证本文中所提出的面形检测方法的正确性,使用MATLAB仿真面形发生变化时用该方法检测的结果。仿真镜面面形方程中X和Y的取值在1pixel~20pixel范围内,以步长为1取值,z=20cosY, 如图 3所示。
图 4a是沿X方向的法线偏差分布图,图 4b是沿Y方向的法线偏差分布图,图 4c是沿Z方向的法线偏差分布图。
Figure 4. a—normal deviation distribution of X direction b—normal deviation distribution of Y direction c—normal deviation distribution of Z direction
表 1和表 2中分别给出了镜面中3个不同位置被测点的理论法线向量和实际法线向量,得出理论法线向量和实际法线向量的差值均值在1mrad以内。
Table 1. Theoretical normal vector results
theoretical normal vector results X direction/rad Y direction/rad Z direction/rad position 1 1.5708 0.3523 1.9231 position 2 1.5708 2.8432 1.8692 position 3 1.5708 2.9259 1.7865 Table 2. Actual normal vector results
actual normal vector results X direction/rad Y direction/rad Z direction/rad position 1 1.5711 0.3528 1.9235 position 2 1.5711 2.8441 1.8701 position 3 1.5713 2.9267 1.7874 -
待测玻璃面板为三星某型号手机玻璃面板,屏幕尺寸为160mm×80mm,投影光屏尺寸为400mm×250mm。实验中相机型号为CatchBest UC320C,分辨率为2048pixel×1536pixel,相元尺寸为3.2μm×3.2μm, 综合考虑视场大小和光屏到被测物的距离,相机采用焦距为25mm的镜头。使用的相机均已经通过MATLAB相机标定工具箱严格标定出内参量,其中焦距为25.63mm,主点坐标为(3.15pixel, 3.09pixel), 2阶径向畸变系数为0.07,4阶径向畸变系数为-2.3,畸变模型为Brown畸变。可通过校准相机参量误差,提高条纹反射系统的测量精度[21]。摄影测量系统采用V-STAR动态摄影测量系统。搭建如图 5所示的实验系统图。
-
测量是在暗室环境下进行的,采集图像之前需要调节相机的曝光时间,使得到的图像不仅具有高对比度而且不会过饱和。图 6a和图 6b为测量相机拍摄的经被测镜反射后的横、竖条纹图像。
对手机玻璃面板在一天内重复测量7次,测量间隔为10min。图 7为其中一次的测量结果。其中图 7a表示的是沿方向的法线偏差分布图,图 7b表示的是沿Y方向的法线偏差分布图,图 7c表示的是沿Z方向的法线偏差分布图。
Figure 7. a—normal deviation distribution in X direction b—normal deviation distribution in Y direction c—normal deviation distribution in Z direction
表 3中给出了被测面上某一个点沿X方向、Y方向、Z方向法线偏差的7次测量结果。表 4中给出被测镜面在X方向、Y方向、Z方向法线偏差的7次测量结果的标准差。
Table 3. Seven measurements at the same point
number of measurements normal deviation of X direction/mrad normal deviation of Y direction/mrad normal deviation of Z direction/mrad 1 0.0886 0.0127 0.0126 2 0.0877 0.0126 0.0126 3 0.0869 0.0123 0.0126 4 0.0844 0.0126 0.0126 5 0.0861 0.0126 0.0125 6 0.0853 0.0127 0.0126 7 0.0847 0.0125 0.0128 standard deviation 0.0016 0.0001 0.0001 由表 3可知, 被测镜面同一点7次重复测量沿X方向、Y方向、Z方向的法线偏差的标准差为0.0016mrad, 0.0001mrad, 0.0001mrad。表 4表明, 7次测量,沿着X方向、Y方向、Z方向的总体法线偏差均在1mrad内,证明测量方案可靠性高。
Table 4. Standard deviation of the spell deviation
results of seven measurements standard deviation of normal deviation of X/mrad standard deviation of normal deviation of Y/mrad standard deviation of normal deviation of Z/mard maximum 0.9899 0.9507 0.8978 minimum 0.8121 0.9182 0.8642 average 0.9112 0.9365 0.8835
基于条纹反射法的平面镜面形测量
Plane mirror shape measurement based on fringe reflection method
-
摘要: 为了评价平面镜面形误差,采用基于条纹反射法的面形检测原理、通过平面镜法线偏移实现面形误差评价的方法。将正弦条纹投射至光屏,经被测平面镜反射至相机,通过小孔成像原理和四步相移法获得被测平面镜上一点的法线,并与经最小二乘算法面形重建的该点理论法线比较,获得该点法线偏移,利用法线偏移实现对平面镜的面形评价。对测量方法进行计算机仿真,以证明其正确性;搭建实验测量系统,并对某一款手机屏幕进行测量实验。结果表明,该测量方法的重复性精度优于1mrad,测量重复性高。该研究为玻璃面板等具有高反射特性的平面面形误差提供了参考。Abstract: In order to evaluate the plane mirror surface shape error, the surface shape detection principle based on the fringe reflection method was adopted, and the method for realizing the plane shape error evaluation by the plane mirror normal offset was used. The sine fringe was projected on the light screen, reflected by the plane mirror to the camera, and the normal of a point on the plane mirror was obtained through the imaging principle of the small hole and the four-step phase shift method. And this point was reconstructed by the least square algorithm surface shape. Theoretical normal comparison was used to obtain the normal offset of the point, and the surface shape evaluation of the plane mirror was achieved by using the normal offset. Computer simulation of the measurement method to prove its correctness; set up an experimental measurement system, and perform measurement experiments on a certain mobile phone screen. The results show that the repeatability accuracy of this measurement method is better than 1mrad, and the measurement repeatability is high. This study provides a reference for the flat surface shape error of glass panels and other products with high reflection characteristics.
-
Key words:
- gratings /
- plane mirror /
- shape detection /
- fringe reflection method /
- normal line deviation
-
Table 1. Theoretical normal vector results
theoretical normal vector results X direction/rad Y direction/rad Z direction/rad position 1 1.5708 0.3523 1.9231 position 2 1.5708 2.8432 1.8692 position 3 1.5708 2.9259 1.7865 Table 2. Actual normal vector results
actual normal vector results X direction/rad Y direction/rad Z direction/rad position 1 1.5711 0.3528 1.9235 position 2 1.5711 2.8441 1.8701 position 3 1.5713 2.9267 1.7874 Table 3. Seven measurements at the same point
number of measurements normal deviation of X direction/mrad normal deviation of Y direction/mrad normal deviation of Z direction/mrad 1 0.0886 0.0127 0.0126 2 0.0877 0.0126 0.0126 3 0.0869 0.0123 0.0126 4 0.0844 0.0126 0.0126 5 0.0861 0.0126 0.0125 6 0.0853 0.0127 0.0126 7 0.0847 0.0125 0.0128 standard deviation 0.0016 0.0001 0.0001 Table 4. Standard deviation of the spell deviation
results of seven measurements standard deviation of normal deviation of X/mrad standard deviation of normal deviation of Y/mrad standard deviation of normal deviation of Z/mard maximum 0.9899 0.9507 0.8978 minimum 0.8121 0.9182 0.8642 average 0.9112 0.9365 0.8835 -
[1] 51TOUCH.TFT-LCD market and development trend of substrate glass[EB/OL].(2018-3-23)[2019-10-10].http://www.51touch.com/material/news/dynamic/2018/0323/49999.html(in Chinese). [2] MORING I, AILISTO H, KOIVUNEN V, et al.Active 3-D vision systom for automatic model-based shape inspection[J].Optics and Lasers in Engineering, 1989, 10(3):149-160. [3] ZHAO X.The research of non-contact measurement technique of the toughened glass's surface topography[D].Taiyuan: North University of China, 2014: 1-5(in Chinese). [4] PIATTI D, RINAUDO F.SR-4000 and CamCube 3.0 time of flight (ToF) Cameras: Tests and comparison[J]. Remote Sens, 2012, 4(4): 1069-1089. doi: 10.3390/rs4041069 [5] HORN B K P. Obtaining shape from shading information[M]. New York, USA:McGraw-Hill, 1989:123-171. [6] YANG P Ch.3-D measurement based on binocular structured light system[D].Hangzhou: Zhejiang University, 2014: 1-3(in Chinese). [7] HORN B K P, BROOKS M J.The variation approach to shape from shading[J]. Computer Vision, Graghics, and Image Processing, 1986, 33:174-208. doi: 10.1016/0734-189X(86)90114-3 [8] ZHAO X.3-D shape measurement of specular surface based on grating progection[D].Hefei: Hefei University of Technology, 2017: 15-20(in Chinese). [9] GENG J.Structured-light 3-D surface imaging: A tutorial[J].Advances in Optics and Photonics, 2011, 3(2): 128-160. doi: 10.1364/AOP.3.000128 [10] BABAIE G, ABOLBASHARI M, FARAHI F.Dynamics range enhancementin digital fringe projection technique[J].Precision Engineering, 2015, 39: 243-251. doi: 10.1016/j.precisioneng.2014.06.007 [11] WADDINGTON C, KOFMAN J. Modified sinusoidal fringe-pattern projection for variable illuminance in phase-shifting three-dimensional surface-shape metrology[J].Optical Engineering, 2014, 53(8): 084109. doi: 10.1117/1.OE.53.8.084109 [12] LIN H, GAO J, MEI Q, et al.Adaptive digital fringe projection technique for high dynamic range three-dimensional shape measurement[J].Optics Express, 2016, 24(7): 7703-7718. doi: 10.1364/OE.24.007703 [13] XIAO J.Study of surface shape measurement of the solar concentrator mirrors in solar thermal power applications[D].Changchun: Changchun Institute of Optic, Fine Mechanics and Physics Chinese Acadmy of Sciences, 2015: 17-36(in Chinese). [14] YAN Q H, LI Y, JIANG Y T, et al. High precision phase estimation of projected fringes in dynamic 3-D measurement[J].Laser Technology, 2019, 43(5): 29-33(in Chinese). [15] DONG F Q, DA F P, HUANG H. Windowed Fourier transform pro-filometry based on advanced S-transform[J].Acta Optica Sinica, 2012, 32(5):512008(in Chinese). doi: 10.3788/AOS201232.0512008 [16] QIAN K M.Two-dimensional windowed Fourier transform for fringepattern analysis: Principles, applications and implementations[J].Optics & Lasers in Engineering, 2007, 45(2):304-317. [17] HARIHARAN H, OREB B F, EIJU T.Digitalphase-shiftininterferometry:A simple error-compensating phase calculation algorithm[J].Applied Optics, 1987, 26(13): 2504-2506. doi: 10.1364/AO.26.002504 [18] SCHWID J.Phase shifting interferometry:reference phase error reduction [J].Applied Optics, 1989, 28(18): 3889-3892. doi: 10.1364/AO.28.003889 [19] FENG G H, ZHANG D Y, WU W Q.Pose estimation of moving object based-on dual quaternion from monocular camera[J].Geomatics and Information Science of Wuhan University, 2010, 35(10):1147-1150(in Chinese). [20] PENG Ch Q, HE Y M, WANG J.A novel high-accuracy deflectometric profiler system based on normal tracing principle[J].Nuclear Techniques, 2017, 40(9):5-14(in Chinese). [21] ZHOU T, CHEN K, WEI H Y, et al.Improved system calibration for specular surface measurement by using reflections from a plane mirror[J].Applied Optics, 2016, 55(25): 7018-7028. doi: 10.1364/AO.55.007018