1.Introduction

A large number of optical measurement techniques have been studied for the three-dimensional(3D)measurement of objects with a diffuse surface[1-3].However,the measurement of specular surfaces remains a challenge.Interferometry is a common method for measuring re flective surfaces with high accuracy;however,it is extremely sensitive to environmental noise[4].In addition,interferometry cannot measure a freeform surface with large gradients(steep sides)due to the limitations of the measurement principle.Stereo deflectometry is an important technique for the measurement of freeform specular surfaces.This technique has been widely investigated recently due to its advantages of non-contact full- field measurement and high accuracy[5-8].Gradient information of the testing surface is obtained by using the geometric relationship of the components in a stereo deflectometry system.The reconstruction shape of the testing object is acquired through an integral calculation of the gradient information[9,10].Therefore,the measurement accuracy of stereo deflectometry is strongly affected by the calculation accuracy of the gradient information and the determination accuracy of the geometric parameters of the components.

（1）色谱条件：XbridgeTM-C18色谱柱（250 mm×4.6 mm，5 μm）；填充剂为十八烷基硅烷键合硅胶；流动相为乙腈-0.5%氨水，梯度洗脱：0～45 min，30%～60%乙腈；45～80 min，60%～80%乙腈；80～120 min，80%乙腈；体积流量0.5 mL/min；检测波长235 nm；柱温30 ℃；进样量20 μL。

A considerable amount of research has been conducted to enhance measurement accuracy in stereo deflectometry.The in fluence of imperfect performance of the components,as compared with an ideal imaging model,has been discussed,and several methods have been investigated to suppress the effects of component imperfections.Petz and Fischer[11]studied the in fluence of the error sources from the display,including form error,the refractions effect,and color display problems.Xu et al.[12]proposed an imaging model and analyzed the in fluence of the system's screen pixel size and the period of fringe displayed on the screen on the sampling phase error and the normal error.Correct phase measurement is crucial in stereo deflectometry measurement techniques.However,error sources such as electronic noise result in random phase error and nonlinear phase error in deflectometry.Yue et al.[13]investigated a technique to eliminate nonlinear phase error in a deflectometry system.Wu et al.[14]studied a phase error reduction technique for deflectometry that included the integration of several error-elimination methods used in fringe projection profilometry,in order to decrease the random noise and nonlinear response noise.Another important aspect of system configuration is system calibration.Standard calibration methods based on special equipment,such as calibration targets[15,16],plane mirrors with special markers[17-19],and extra displays or cameras[20,21],have been developed for deflectometry calibration.However,there may be an introduction of new error sources from the calibration equipment.Therefore,Werling[22]and Xiao et al.[23]explored a calibration method to calculate the geometric relative relation of the units in deflectometry by only applying a plane mirror without markers.Conventional calibration methods[24-28]for deflectometry complete camera calibration and systemic calibration separately.Camera calibration is conducted first,and systemic geometric parameters are then calculated based on the camera calibration result.However,the objective functions of parameter optimization during camera calibration and systemic calibration are different,which results in an inconsistency when combining all systemic optics parameters.In addition,the camera calibration error is propagated to the systemic calibration,which leads to larger systematic deviations.In order to eliminate the error propagation and the inconsistency in different calibration steps,holistic calibration techniques have recently been investigated.For example,self-calibration methods for a stereo deflectometry system were studied by Olesch et al.[29]and Faber et al.[30,31],and Ren et al.[32]researched an iterative optimization algorithm for the calibration of a stereo deflectometry system.Based on an iterative distortion compensation algorithm,Xu et al.[33]investigated a calibration approach for stereo deflectometry.In addition to the abovementioned error sources,arrangements of the components of a deflectometry system affect the system performance.A performance analysis of geometric parameters is essential for any optical measurement system based on the triangulation principle.Zhang et al.[34]studied the performance analysis of a 3D measurement system based on fringe projection,and Zhao et al.[35]analyzed the in fluence of four system geometric parameters on a direct phase measuring deflectometry(DPMD)system.Although these approaches involve the development and optimization of system geometric parameters,the methods and research in these studies do not sufficiently cover stereo deflectometry.

In order to improve measurement accuracy,this paper presents a novel geometric parameters analysis in a stereo deflectometry system.Fourgeometric parameters ofstereo deflectometry systems are analyzed:the angle between the measured surface normal and the camera ray,the distance between the measured surface and the camera,the distance between the measured surface and the screen,and the angle between the cameras.The in fluence of these geometric parameters on the measurement accuracy is reported.

2.Principle and analysis

A stereo deflectometry system is typically composed of a liquid crystal display(LCD)screen displaying coded patterns and two charge-coupled device(CCD)cameras capturing the re flected patterns through the re flection of the measured surface.The measurement principle of stereo deflectometry is demonstrated in Fig.1.The screen displays phase-shifting sinusoidal fringe patterns in sequence.The cameras capture the patterns re flected by the measured surface synchronously.Because the displayed patterns are mutually perpendicular,two orthogonal absolute phase maps in terms of each camera can be acquired by using phase wrapping and unwrapping methods.For an arbitrary point(S1,S2,S3,...)in space,its image on the camera's image plane can be calculated based on camera calibration.Its corresponding physical point on the fringe-displaying screen can be obtained by using the relation between the absolute phase value and the physical coordinate on the screen.Therefore,the equivalent normal of the point in space can be acquired from a triangular geometric relationship composed of the point in space,the corresponding point on the camera image plane,and the corresponding point on the screen.During the measurement process,one camera acts as the main camera.Points in space are searched for along the camera rays of the main camera.The other camera is a reference camera;the data from this camera may be used to define the position of the searched points in space.Since normal vectors of the surface point calculated from the main camera and reference camera should be overlapped,initial 3D data with gradient information of the test surface can be discovered by searching for points in space and matching the normal vectors from the cameras.The form of the measured surface is reconstructed by integrating the acquired gradient information[9,10].

从表2可以看出，国内近5年来，对于英语词汇学习与其他变量的关系研究最多，其次是对教学方面的研究。其中最少的是文献综述的研究，一方面可能是本身研究的文献数不多，另一方面可能是研究内容具有局限性，重视度不够。对于变量关系的研究反映了我国研究者对影响英语词汇学习因素的重视，研究面相对较广。除此之外，对于教学和学习的直接研究一直较受重视，不可忽视的是对其他方面的研究也或多或少地给词汇学习和教学提供了一定的启示。

2.1.The in fluence ofγ

Fig.1.Measurement principle of stereo deflectometry.

Fig.2.Arrangement of stereo deflectometry.O1 and O2 are the optical centers of the main camera and the reference camera,respectively.

Through the re flection of the measured surface,the image of the displaying screen can be captured directly by the camera in a deflectometry system.The relative positions of the imaging plane of the camera,the measured surface,and the mirrored screen are depicted in Fig.5.Based on a pinhole model,a point on the mirrored screen is captured by the camera.I is the image of on the imaging plane,andand δI denote the location uncertainty caused by phase error on the mirrored screen and imaging plane,respectively.When the measured surface is located at Surface 1,the surface is perpendicular to the camera ray. results in an uncertainty of δS1 on the measured surface.In contrast,when the measured surface is located at Surface 2,whereγis not equal to 0,the uncertainty δS2 caused byon the measured surface can be described as follows:

The uncertainty δZ along the Z direction can be calculated as follows:

In addition to image blurring brought about by defocusing,curvature of the measured surface blurs the mirrored fringe patterns.In order to demonstrate this point,an experiment was carried out to capture the fringe patterns re flected from mirrors with different curvatures.Three mirrors were tested:a flat mirror,a concave mirror with a 200 mm radius of curvature(f),and a concave mirror with a 100 mm radius of curvature.Fig.13 shows the captured fringe patterns when L s is 80 and 120 mm.The results demonstrate that with increasing mirror curvature,fringe density decreases and blurring increases.Increasing L s will amplify the in fluence of the mirror's curvature.Therefore,L s should be around 80 mm when comprehensively considering the measurement accuracy and measurement range of the mirror curvature.

2.2.The in fluence of L s

In order to illustrate the in fluence of L s,the screen of a stereo deflectometry system is first placed at D,and then moved to D′,as shown in Fig.6.I is the image of a point P on the screen through the re flection of the measured surface.The normal n of the surface can be calculated based on the incident ray PS and the re flection ray SI,according to the law of re flection.Because of the in fluence of phase error,there is a location uncertainty δP for P.n1 and n2 are the calculated normal values affected by δP when the screen is located at D and D′,respectively.α1 denotes the angle difference between n1 and n,andα2 represents the angle difference between n2 and n.Since L s is much larger than δP,the value ofα1 and α2 can be calculated according to Eq.(3):

Fig.3.Relationship between geometric parameters and error sources.

Fig.4.Error analysis procedures.

where α is the normal uncertainty caused by δP.It is clear that α decreases with the increase of L s,according to Eq.(3).However,other factors limit the length of L s in a stereo deflectometry system.First,because the cameras in stereo deflectometry are required to focus on the measured surface,the fringe patterns on the screen are defocused due to the length of L s;this results in the increase of δP and α.Second,the curvature of the measured surface blurs the mirrored screen.The increase of L s increases the degree of fuzziness,which leads to the increase of αas well.

2.3.The in fluence of L d

3.1.4.Verification of the analysis of θ

The relationship between θ and the measurement error was studied through a simulation in which L s,L d,andγwere set as 80 mm,300 mm,and 0°,respectively.Fig.10(d)shows the experimental result,which shows that the measurement error decreases gradually with increasing θ,and is almost stable from 10°on.This decreasing trend verifies the analysis in Section 2.4.Therefore,in general,θ should be no less 10°.

以聚苯乙烯泡沫板第1次试验为例，计算K值.热流系数f1=0.492 W/K，f2=5.483 W/K，δ=0.045 m，A=3.24 m2，其余参数取值见表1、表3.

以马其顿为例，该国的地震区划图采取了1 000年内超越概率为63%的最大地震烈度作为抗震设防的基本烈度，将马其顿全国划分为7、8、9三个设防烈度区[6]。按照我国相关抗震设防的规定，三个烈度即众值烈度(50 a内超越概率63%)、基本烈度(50 a内超越概率10%)、罕遇烈度(50 a内超越概率2%～3%)的相关定义[7]，马其顿抗震设防基本烈度概率接近于罕遇烈度，根据《中国地震动参数区划图贯宣材料》所述，罕遇烈度相对于众值烈度比值的优势分布为3倍，罕遇烈度对基本烈度比值的优势分布为1.8倍的情况，马其顿相应抗震设防烈度区7、8、9度区基本对应于中国规范的6、7、8度区。

Therefore,the in fluence of the change in L d on δS2 and δZ can be neglected.

2.4.The in fluence of θ

The in fluence of θ is depicted in Fig.8.The main camera is placed at plane MC,and the reference camera is first tested at plane RC and then moved to plane RC′in order to investigate the in fluence of θ.O1,O2,and are the optical centers of MC,RC,and RC′,respectively.S is a point on the measured surface and I1 is the image of S on MC.denotes the intersection of the camera ray I1S and the mirrored screen,and P1 represents the actual point on the screen of.The normal n of S can be calculated according to the triangular relationship composed of P1,S,and I1,according to the law of re flection.Based on the same principle,the normal vectors of S′and S′′,which are two points on the camera ray of I1S,can be calculated.Fig.9(a)provides an enlarged view of S′.n11,n21,and n31 denote the calculated normal vectors of S′based on MC,RC,and RC′,respectively.Similarly,S′′is enlarged in Fig.9(b).The calculated normal vectors of S′′from MC,RC,and RC′are denoted as n12,n22,and n32.Phase error leads to a measurement uncertainty for each calculated normal vector of δ11,δ21,δ31,δ12,δ22,and δ32,as shown in Fig.9.It is clear that the calculated normal vectors from MC and RC are overlapped.Therefore,points between S′and S′′would be wrongly judged to belong to the measured surface when using MC and RC.In contrast,there is no overlapping between the calculated normal vectors based on MC and RC′.S′and S′′can be determined correctly as not being points on the measured surface.The above analysis indicates that increasing θ can increase the system measurement accuracy.

Fig.5.Relative positions of the imaging plane of the camera,the measured surface,and the mirrored screen.

Fig.6.Illustration of the in fluence of L s.S is a point on the measured surface.

Fig.7.Illustration of the in fluence of L d.

3.Experiments

3.1.Simulation

3.1.1.Verification of the analysis ofγ

绛蜡丛中一讲庐，披图想见过庭趋。埙篪韵事真馨逸，得许升堂展谒无。梓桥俯仰傍寒林，伴结幽芳夙抱深。家法岂惟传治谱，先将铁石炼胸襟。

Four simulations were performed to test the proposed performance analysis.The settings of the analyzed geometric parameters in the simulations are listed in Table 1.The simulated main camera and reference camera both have a resolution of 1616×1216 pixels.The pixel pitch of the cameras is 0.0035 mm.The fringe-displaying screen in the simulation has a resolution of 1024×1024 pixels and a pixel pitch of 0.294 mm.Through the re flection of a simulated flat specular surface,sinusoidal fringe patterns whose period occupies eight pixels of the screen were displayed in turn and captured by the cameras simultaneously.In order to simulate the actual measurement environment,random errors were added into the physical locations of the screen and the camera pixel location based on general experimental results[12,32,33].

Fig.8.Illustration of the in fluence of θ.P21,P22,P31,and P32 are points on the screen.P′ 21,,,andare the points on the mirrored screen.I21,I22,I31,and I32 are image points on the CCD planes.

Fig.9.Enlarged view of the calculated normal values.(a)Enlarged view of S′;(b)enlarged view of S′′.

A simulation was performed to test the in fluence ofγ.In the simulation,L d,L s,and θ were set as 300 mm,80 mm,and 10°,respectively.Random errors with a maximum value of 0.0178 mm and 0.05 pixels were added into the physical locations of the screen and camera pixel,respectively.γvaried from 0°to 11°in increments of 2°or 3°.The relationship between γand the measurement error was acquired,as shown in Fig.10(a).

Fig.10(a)shows that with increasingγ,the measurement error increases gradually.This increasing trend verifies the analysis in Section 2.1.Therefore, γshould theoretically be 0°.However,because the fringe patterns on the screen must be captured by the cameras through the re flection of the surface,γcannot be 0°in an actual stereo deflectometry system.In general,γshould be as small as possible.

3.1.2.Verification of the analysis of L s

L s varies from 40 to 200 mm in increments of 40 mm.According to the results of the actual experiment,random errors with maximum values of 0.0170,0.0178,0.022,0.033,and 0.051 mm were added to the physical locations of the screen when L s was 40,80,120,160,and 200 mm,respectively.Random errors with a maximum value of 0.05 pixels were added to the camera pixel location.The relationship between L s and measurement error is shown in Fig.10(b).

在石英岩、片麻岩、片岩、变粒岩中，石榴石广泛发育，十字石常与蓝晶石和石榴石共生；黑云变粒岩及黑云片麻岩中常可见到矽线石；蓝晶十字黑云变粒岩、黑云母石英片岩及黑云母片岩中均有堇青石出现，这些都是典型变质特征矿物，反映变质相为中低变质相，即角闪岩相。

Fig.10(b)shows that the measurement error decreases dramatically when L s increases from 40 to 120 mm.The main reason for this decreasing trend is that the gradient uncertainty caused by location uncertainty on the screen decreases with increasing L s,according to Eq.(3).However,because phase errors are magnified with the increase of L s,the measurement error increases slightly when L s changes from 120 to 200 mm.

Three principle error sources directly affect the measurement accuracy of a stereo deflectometry system:calibration error,phase error,and gradient calculation error.Other error sources impact the measurement accuracy by acting on these three error sources.Fig.2 shows the arrangement of the relevant components in a stereo deflectometry system.In this figure,L d is the distance between the main camera and the measured object surface,γ denotes the angle between the main camera ray and the surface normal n,and L s represents the distance between the fringedisplaying screen and the measured object surface.The angle between the main camera and the reference camera is denoted as θ.The geometric parameters γ,L s,L d,and θ are analyzed for their in fluence on the three error sources and overall measurement accuracy.These four parameters are independent of each other in geometry.Fig.3 shows the relationship between the geometric parameters and the error sources.In this figure,the arrows indicate that the factor at the tail of the arrow affects the error source at the arrow head.The geometric parameters affect gradient calculation accuracy independently.Stereo deflectometry measures the form data of a specular surface based on the relative relation between the cameras and the fringe-displaying screen in the system.It is important to perform a calibration process to determine the location of the components in the system.In Fig.3,the phrase‘‘calibration accuracy”refers to the accuracy of the calibration process.Calibration accuracy is affected by phase error[33].Because the cameras in the system focus on the measured object,the captured fringe patterns displayed on the screen are defocused.L s affects the degree of defocus of the fringe patterns,and leads to a change in phase error.Hence,L s has an in fluence on calibration accuracy.Although the distance between the screen and the object also impacts the gradient calculation accuracy through its in fluence on the degree of defocus of the fringe patterns,the impact from phase accuracy is independent of the in fluence caused by other geometric parameters.Therefore,the in fluences of the geometric parameters on the measurement accuracy are not coupled.Fig.4 provides an overall flowchart to clarify the error analysis procedure.Four error models are investigated to analyze the in fluence of the geometric parameters.The conclusions of the analysis are then verified through simulation and actual experimentation by modifying the measuring conditions.Finally,an appropriate combination of geometric parameters is obtained.

3.1.3.Verification of the analysis of L d

L s,θ,and γwere set as 80 mm,10°,and 0°,respectively,in a simulation to study L d.Fig.10(c)shows the relationship between L d and the measurement error as L d increased from 200 to 700 mm in increments of 50 mm.

妇人与少女都越走越远了，我仍然站在原地，想着时光怎样改变人的心和人的面貌。想着二十年的岁月可以有这样剧烈的改变，这样遥远的差异，不禁怅然。

The simulation result shows that with increasing L d,the measurement error decreases gradually.This decreasing trend confirms the analysis in Section 2.3.Hence,theoretically,L d should be as large as possible.However,a large L d results in a similarly large size of the stereo deflectometry system.Thus,in general,L d should be determined based on an overall consideration of the measurement accuracy and the application.

A camera is placed at C and C′successively to analyze the in fluence of L d,as shown in Fig.7.O1 andare the optical centers of C and C′,respectively.The location uncertainty of I is denoted as δI due to the in fluence of phase error.n1 and n2 are the calculated normal values based on C and C′,respectively,considering the in fluence of δI.The calculation errors of n1 and n2 are represented as β1 and β2.Since L d is much larger than δI, β1 and β2 can be calculated using Eq.(4):

where β is the normal uncertainty caused by δI.On the one hand,β decreases with an increase of L d,according to Eq.(4).On the other hand,based on the analysis of Eqs.(1)and(2),increasing L d enlarges δS2 and δZ,which leads to measurement error.However,if the camera has a long focus length,L d is far greater than L s,and Eqs.(1)and(2)can be simplified to Eqs.(5)and(6):

3.2.Actual experiments

Based on the above analysis and simulation results,a measurement system was set up,as shown in Fig.11.Two Lumenera CCD sensors(Model Lw235M)were used as the main camera and reference camera of the system.The resolution of the sensors was 1616×1216 pixels.Xu et al.[12]have determined that decreasing the pixel size of the fringe-displaying screen can increase measurement accuracy.The iPad Pro[36]has an ultra fine pixel size of 0.096 mm,which is much smaller than the most common display pixel size(around 0.3 mm).Therefore,under the control of DisplayDuet[37],an iPad Pro with a size of 12.9 in(1 in=2.54 cm)was used as the fringe-displaying screen.The system was calibrated with the method proposed in Ref.[33].

Table 1 Parameter settings in simulations.

γ (°) L s(mm) L d(mm)θ (°)Simulation 1 0,2,4,6,8,11 80 300 10 Simulation 2 0 40,80,120,160,200 300 10 Simulation 3 0 80 200,250,300,350,400,450,500,550,600,650,700 10 Simulation 4 0 80 300 1,1.9,5.7,8.6,11.6,15.6

Fig.10.The obtained relation between geometric parameters and measurement error based on simulation results.(a)Relation betweenγand measurement error;(b)relation between L s and measurement error;(c)relation between L d and measurement error;(d)relation between θ and measurement error.

Fig.11.Setup of the stereo deflectometry system.

First,experiments were conducted to investigate the in fluence of L s on phase error.Since the cameras in a stereo deflectometry system are required to focus on the measured surface,L s causes the mirrored screen to be out of the depth of focus of the cameras,and therefore blurs the captured fringe patterns.The phase error increases with an increase of the degree of fuzziness,which leads to a decrease in measurement accuracy.

The relation between L s and the phase error was investigated through actual experimentation.The errors of absolute phase maps can be obtained when L s varies from 0 to 200 mm in increments of 40 mm by applying phase-shifting and phase-unwrapping techniques[38,39].Fig.12 shows the experimental result.When L s is 0 mm,the phase error is 0.0415 rad,which is significantly larger than the phase error(0.0151 rad)obtained when L s is 40 mm.The reason for this phenomenon is that when the camera is focused on the mirrored screen,LCD pixel grids and a moiré fringe[40]are recorded by the camera.The recorded pixel grids and moiré fringe generate large errors on the phase maps.When L s is greater than 40 mm,the phase error increases dramatically with increasing L s.The main reason for this increasing trend is that the degree of defocus of the mirrored screen is magnified with an increase in L s.The experimental results were used to provide the correct phase errors for the abovementioned simulation study of L s.

传统的干法回收处理LFP动力电池成本高、利润低，学者们开发出了新的干法回收技术。本课题组在空气中高温处理正极片后，将 LiFePO4氧化为 Li3Fe2(PO4)3及 Fe2O3并作为再生反应原料，加入适量还原剂，650 ～ 750℃高温碳热还原再生LiFePO4，获得纯相的再生LiFePO4/C材料，如图5。

Fig.12.Relation between L s and phase error.

According to Eqs.(1)and(2),it is obvious that δS2 and δZ increase with increasingγ.

Fig.13.Relation between mirrored fringe pattern and mirror curvature.(a)Flat mirror when L s is 80 mm;(b)concave mirror with 200 mm radius of curvature when L s is 80 mm;(c)concave mirror with 100 mm radius of curvature when L s is 80 mm;(d) flat mirror when L s is 120 mm;(e)concave mirror with 200 mm radius of curvature when L s is 120 mm;(f)concave mirror with 100 mm radius of curvature when L s is 120 mm.

Fig.14.Obtained relation between geometric parameters and measurement error based on actual experimental results.(a)Relation betweenγand measurement error;(b)relation between L s and measurement error;(c)relation between L d and measurement error;(d)relation between θ and measurement error.

Fig.15.Measurement result of a stock concave mirror.(a)Captured image of a concave mirror;(b)obtained form of the mirror;(c)measurement difference.

An additional experiment was performed using an optical flat with λ/20 flatness from Edmund[41].In the experiment,γwas changed to the following four values:10°,15°,20°,and 25°.Fig.14(a)shows that the measurement errors increase significantly with increasingγ.This result confirms the results of the analysis described earlier.Fig.14(b)shows that the measurement errors have a similar trend to the simulation results when L s is altered to 60,80,100,and 120 mm.Fig.14(c)shows the in fluence of L d on the measurement accuracy.When L d is increased from 250 to 320 mm,the measurement errors decrease gradually.This result corresponds to the analysis above of L d.Fig.14(d)shows the trend in measurement errors for four different values of θ:10°,15°,20°,and 25°.The measurement errors change slightly from 10°,which corresponds to the trend shown in the simulation.It is difficult to conduct a quantitative comparison between experimental results and simulation results because the experimental conditions in an actual experiment,such as the degree of defocus of the cameras and the measurement field,cannot be simulated accurately.The difference in experimental conditions will affect the measurement result.Therefore,the proposed performance analyses were qualitatively tested by comparing the trend of the measurement results.

由于阀门CV值过大，开度变小，黑水流经角阀处形成偏流，从而导致闪蒸，角阀阀芯、阀座出现局部冲刷磨损严重现象。针对该现象应适当降低阀的压差Δp，由流量公式可知，当Δp减小时，qm也减小，为保证通过调节阀的流量不变，就要增大阀门的开度，这样可避免阀门在小开度下工作。解决该问题方案如下：

保持体形，是咱们女人一生的事业。刚生完孩子根本不敢上秤，不想被贴上油腻妇女的标签，就要“管住嘴，迈开腿”。

The geometric parameters of the measurement system were determined according to the proposed performance analysis.θ was set at 10°and L d was set at 300 mm in consideration of the size limit of the measurement system.L s was chosen to be 80 mm.γ was designed to be 10°in order to ensure that the displayed patterns could be ‘‘seen”by the cameras.In order to quantitatively evaluate the measurement accuracy of the stereo deflectometry system,a concave mirror(stock#40-913)from Edmund with λ/8 surface accuracy was measured.Fig.15(a)shows one of the images captured from the main camera with deformed fringe information.The diameter of the measurement field of the system is about 2 in.The test mirror was larger than the measurement field;therefore,the central section of the mirror was measured.Fig.15(b)shows the measured form of the test mirror.The peak value(PV)of the difference between the measurement result and the surface parameters provided by the manufacturer was 154.2 nm,as shown in Fig.15(c).These measurement results demonstrate that a measurement system design based on the proposed performance analysis can acquire the form of a specular surface with high precision.

4.Conclusions

In order to improve the measurement accuracy of stereo deflectometry,the in fluence of four geometric parameters was investigated through error model analysis,simulation testing,and experimental verification.This performance analysis will optimize the design of a stereo deflectometry system and decrease the measurement error caused by system component positioning error.

Future work will include applying the designed stereo deflectometry system to industrial applications,such as by embedding stereo deflectometry into an ultra-precision machining system in order to achieve online measurement,and using stereo deflectometry in vehicle detection for the inspection of defects in car windows and bodies.

Acknowledgements

The authors gratefully acknowledge the UK's Engineering and Physical Sciences Research Council(EPSRC)funding of the EPSRC Centre for Innovative Manufacturing in Advanced Metrology(EP/I033424/1),the EPSRC Future Advanced Metrology Hub(EP/P006930/1),and the EPSRC funding(EP/K018345/1).

Compliance with ethics guidelines

Yongjia Xu,Feng Gao,and Xiangqian Jiang declare that they have no conflict of interest or financial conflicts to disclose.

References

[1]Zhang Z.Review of single-shot 3D shape measurement by phase calculationbased fringe projection techniques.Opt Lasers Eng 2012;50(8):1097-106.

[2]Zuo C,Huang L,Zhang M,Chen Q,Asundi A.Temporal phase unwrapping algorithms for fringe projection profilomery:a comparative review.Opt Lasers Eng 2016;85:84-103.

[3]Zuo C,Chen Q,Gu G,Feng S,Feng F.High-speed three-dimensional profilometry for multiple objects with complex shapes.Opt Express 2012;20(17):19493-510.

[4]Jiang X,Wang K,Gao F,Muhamedsalih H.Fast surface measurement using wavelength scanning interferometry with compensation of environmental noise.Appl Opt 2010;49(15):2903-9.

[5]Knauer MC,Kaminski J,Hausler G.Phase measuring deflectometry:a new approach to measure specular free-form surfaces.In:Osten W,Takeda M,editors.Optical metrology in production engineering.Proceedings of Photonics Europe;2004;Strasbourg,France.Wales:SPIE;2004.p.366-76.

[6]Balzer J,Hofer S,Beyerer J.Multiview specular stereo reconstruction of large mirror surfaces.In:Proceedings of 2011 IEEE Conference on Computer Vision and Pattern Recognition;2011 Jun 20-25;Providence,RI,USA.New York:IEEE;2011.p.2537-44.

[7]Fischer Petz M,Tutsch R.Model-based deflectometric measurement of transparent objects.In:Osten W,editor.Fringe 2013.Proceedings of the 7th International Workshop on Advanced Optical Imaging and Metrology;2014 Jan 1;Nurtingen,Germany.Berlin:Springer;2014.p.573-6.

[8]Kaminski J,Lowitzsch S,Knauer M,Häusler G.Full- field shape measurement of specular surfaces.In:Osten W,editor.Fringe 2005:full- field shape measurement of specular surfaces.Berlin:Springer;2006.p.372-9.

[9]Huang L,Asundi A.Improvement of least-squares integration method with iterative compensations in fringe re flectometry.Appl Opt 2012;51(31):7459-65.

[10]Ren H,Gao F,Jiang X.Least-squares method for data reconstruction from gradient data in deflectometry.Appl Opt 2016;55(22):6052-9.

[11]Petz M,Fischer M.Systematic errors in deflectometry induced by use of liquid crystal displays as reference structure.In:Proceedings of the 21st IMEKO TC2 Symposium on Photonics in Measurement;2013 Sep 16-18;Gdansk,Poland;2013.p.16-8.

[12]Xu Y,Gao F,Zhang Z,Jiang X.Enhancement of measurement accuracy of optical stereo deflectometry based on imaging model analysis.Opt Lasers Eng 2018;111:1-7.

[13]Yue H,Wu Y,Zhao B,Ou Z,Liu Y,Liu Y.A carrier removal method in phase measuring deflectometry based on the analytical carrier phase description.Opt Express 2013;21(19):21756-65.

[14]Wu Y,Yue H,Yi J,Li M,Liu Y.Phase error analysis and reduction in phase measuring deflectometry.Opt Eng 2015;54(6):064103.

[15]Soumelidis A,Fazekas Z,Bodis-Szomoru A,Schipp F,Csakany B,Nemeth J.Specular surface reconstruction method for multi-camera corneal topographer arrangements.Recent Adv Biomed Eng 2009:639-60.

[16]Huang L,Ng CS,Asundi AK.Dynamic three-dimensional sensing for specular surface with monoscopic fringe re flectometry.Opt Express 2011;19(13):12809-14.

[17]Huang L,Xue J,Gao B,McPherson C,Beverage J,Idir M.Modal phase measuring deflectometry.Opt Express 2016;24(21):24649-64.

[18]Breitbarth M,Kühmstedt P,Notni G.Calibration of a combined system with phase measuring deflectometry and fringe projection.In:Lehmann PH,editor.Optical measurement systems for industrial inspection VI.Proceedings of SPIE Europe Optical Metrology;2009;Munich,Germany.Wales:SPIE;2009.p.738909.

[19]Balzer J,Werling S.Principles of shape from specular re flection.Measurement 2010;43(10):1305-17.

[20]Bonfort T,Sturm P.Voxel carving for specular surfaces.In:Proceedings of the Ninth IEEE International Conference on Computer Vision;2003 Oct 13-16;Nice,France.New York:IEEE;2003.p.591-6.

[21]Xiao Y,Su X,You Z.Pose transfer geometrical calibration for fringe-re flection optical three-dimensional measurement.Opt Commun 2013;305:143-6.

[22]Werling S.De flectometry for automatic visual inspection and reconstruction of re flective surfaces[dissertation].Baden-Württemberg:Karlsruhe Institute of Technology;2011.German.

[23]Xiao Y,Su X,Chen W.Flexible geometrical calibration for fringe-re flection 3D measurement.Opt Lett 2012;37(4):620-2.

[24]Zhao W,Su X,Liu Y,Zhang Q.Testing an aspheric mirror based on phase measuring deflectometry.Opt Eng 2009;48(10):103603-5.

[25]Huang L,Ng CS,Asundi AK.Fast full- field out-of-plane deformation measurement using fringe re flectometry.Opt Lasers Eng 2012;50(4):529-33.

[26]Xiao Y,Su X,Chen W,Liu Y.Three-dimensional shape measurement of aspheric mirrors with fringe re flection photogrammetry.Appl Opt 2012;51(4):457-64.

[27]Maldonado AV,Su P,Burge JH.Development of a portable deflectometry system for high spatial resolution surface measurements.Appl Opt 2014;53(18):4023-32.

[28]Wedowski RD,Atkinson GA,Smith ML,Smith LN.Dynamic deflectometry:a novel approach for the on-line reconstruction of specular freeform surfaces.Opt Lasers Eng 2012;50(12):529-33.

[29]Olesch E,Faber C,Hausler G.De flectometric self-calibration for arbitrary specular surfaces.In:Proceedings of the 2011 Annual Conference of the German Society for Applied Optics;2011 Jun 15-18;Ilmenau,Germany;2011.

[30]Faber C,Olesch E,Krobot R,Häusler G.De flectometry challenges interferometry:the competition gets tougher!In:Schmit J,Creath K,Towers CE,Burke J,editors.Interferometry XVI:techniques and analysis.Proceedings of SPIE;2012 Aug 12-16;San Diego,CA,USA.Wales:SPIE;2012.

[31]Faber C.New methods and advances in deflectometry.Erlangen:Erlangen Scientific Press;2012.

[32]Ren H,Gao F,Jiang X.Iterative optimization calibration method for stereo deflectometry.Opt Express 2015;23(17):22060-8.

[33]Xu Y,Gao F,Zhang Z,Jiang X.A holistic calibration method with iterative distortion compensation for stereo deflectometry. Opt Lasers Eng 2018;106:111-8.

[34]Zhang Z,Zhang D,Peng X.Performance analysis of a 3D full- field sensor based on fringe projection.Opt Lasers Eng 2004;42(3):341-53.

[35]Zhao P,Gao N,Zhang Z,Gao F,Jiang X.Performance analysis and evaluation of direct phase measuring deflectometry.Opt Lasers Eng 2018;103:24-33.

[36]iPad Pro[Internet].Hollyhill:Apple,Inc.;c2018[cited 2018 Sep 14].Available from:https://www.apple.com/uk/ipad-pro/.

[37]Duet display[Internet].Duet,Inc.;c2018[cited 2018 Sep 14].Available from:https://www.duetdisplay.com/.

[38]Towers CE,Towers DP,Jones J.Absolute fringe order calculation using optimised multi-frequency selection in full- field profilometry.Opt Lasers Eng 2005;43(7):788-800.

[39]Zhang Z,Towers CE,Towers DP.Time efficient color fringe projection system for 3D shape and color using optimum 3-frequency selection.Opt Express 2006;14(14):6444-55.

[40]Xu Y,Gao F,Ren H,Zhang Z,Jiang X.An iterative distortion compensation algorithm for camera calibration based on phase target.Sensors(Basel)2017;17(6):1188.

[41]Edmund optics[Internet].Illinois:The future depends on optics,Inc.;c2018[cited 2018 Sep 14].Available from:https://www.edmundoptics.com/.