双平面镜双相机数字图像相关技术在三维重建中的应用【增强内容出版】
Digital image correlation (DIC) technology is a processing method commonly employed for image matching, and meanwhile after nearly forty years of development, its accuracy, efficiency, practicality, and other aspects have yielded significant improvement. With the development and progress of science and technology, DIC technology for three-dimensional (3D) measurement should also be economical and practical, with the utilization of relatively simple devices for a full range of functions. DIC measurement systems with the assistance of the camera and external equipment can also be realized with multiple viewpoints and multi-directional measurements, on which many scholars have carried out thorough research. Among them, the single camera system has a flexible field of view adjustment and simple optical path, but features poor stability and low accuracy. The multi-camera system requires the adoption of multiple cameras, and the calibration process is complicated. Although the multi-camera measurement system can improve the range and accuracy of 3D measurement, it is difficult to be widely leveraged in 3D full-field measurement due to the high requirements of environmental conditions and the expensive cameras. Given the shortcomings of the existing single-camera and multi-camera systems, we propose a dual-camera 3D reconstruction method assisted by dual plane-mirror.
We put forward a visual 3D measurement method assisted by a dual plane-mirror, which is to analyze the virtual and real correspondences of the corner points in the mirror surface and thus obtain the reflection transformation matrix. Meanwhile, the virtual and real transformations of the object surface are completed by the reflection transformation matrix, and the 3D full-field measurements are realized finally. Additionally, this method avoids spraying diffuse spots on mirrors to take up the spatial resolution of the camera, making the solution of the reflection transformation matrix easy and efficient. Firstly, the image information of the front surface and the back side surface of the object is acquired simultaneously by the dual-camera DIC measurement system (Fig. 1). Secondly, the calibration plate is placed in front of the plane mirror, and the dual cameras can observe the real calibration plate and the virtual image in the mirror at the same time (Fig. 4). The midpoint method based on the common vertical line is adopted to determine the 3D coordinates of the corner points in space (Fig. 5), and the specific positions of the dual plane mirrors are finally determined by changing the position of the calibration plate several times. Finally, the reflection transformation matrix is solved by the mirror position equation, and then the 3D reconstruction of the object is completed.
To verify the accuracy of the proposed method, we conduct static and dynamic experiments on the measured parts respectively. In the static experiments, the 3D contour of the game coin is reconstructed, and in the dynamic experiments, the thermal deformation of the five-side aluminum column is investigated (Fig. 6). By employing the proposed method, the dual-mirror equation and reflection transformation matrix can be obtained under the mirror angle of 108° (Table 1). The front and back contours of the ordinary game coin are reconstructed in three dimensions, the theoretical thickness of the game coin is 1.80 mm, and the measured thickness is around 1.75 mm (Fig. 9). The reconstruction method of spraying diffuse spots on the mirror surface is compared to verify the 3D reconstruction accuracy of the proposed method (Fig. 9), and the 3D reconstruction of the game coin by the proposed method is found to be better than that of spraying diffuse spots on the mirror surface. Meanwhile, the proposed method avoids taking up the spatial resolution of the camera by spraying diffuse spots on the mirror surface, with higher accuracy.
Aluminum column 3D reconstruction and thermal deformation measurement results are shown. Firstly, the reconstruction results of the surface profile of the five-side aluminum column are obtained by the proposed method (Fig. 10), the real height of the aluminum column is 70.00 mm±0.01 mm, and the average measurement height is 70.0035 mm, which is a sound measurement effect. Secondly, the average height change of the five outer surfaces of the aluminum column can be obtained during thermal deformation (Table 2). The thermal deformation displacement cloud map of the outer surfaces is shown in Fig. 11 and the height change of different surfaces in the cooling process is illustrated in Fig. 12. To more intuitively demonstrate the accuracy of the proposed method of real-virtual transformation, we compare and analyze the deviation values of the height change obtained by the two methods at each node (Fig. 13), which shows that the proposed method has higher measurement accuracy.
We propose a dual plane mirror-assisted visual DIC 3D full-field measurement method. The static experiment results indicate that the proposed method is better than the reconstruction method of spraying diffuse spots on mirrors for the 3D reconstruction of game coins with higher accuracy. The results of dynamic thermal deformation experiments indicate that when the temperature of the five-side aluminum column is reduced from 330 to 20 ℃, the height change of the outer surfaces of the column is basically consistent with the simulation results of the finite element software, and the deviation values of the height change measured by the proposed method are smaller than those of the method of spraying diffuse spots on mirrors. Since the proposed method can avoid spraying diffuse spots on mirrors to take up the spatial resolution of the camera, it features simple operation, high measurement accuracy, and sound application perspectives.
1 引言
数字图像相关(DIC)技术是一种图像匹配的常用处理方法,经过近四十年的发展[1-3],在精度、效率、实用性等方面都取得了显著的提高,并且在土木工程、生物组织、航空航天等领域都具有很高的实用价值[4]。DIC测量系统借助于相机和外部设备可以实现多视角、多方位的测量[5-7]。
在三维(3D)测量方面[8-12]很多学者开展了深入研究。2011年,Orteu等[13]开发了第一个“主摄像机”结构的多视角数字图像相关测量系统,该系统使用四台相机进行匹配,实现被测件形状的三维测量。2016年,Li等[14]提出了一种用于全场厚度应变测量的多相机DIC测量系统,该系统也使用了四个相机,两对相机分别测量试件的正面和背面,每对相机构成一个双目DIC测量系统,通过标定将两个子系统联系起来,实现全场厚度应变测量。随后,还有学者提出使用更多的相机建立用于表面形变测量的多视角DIC测量系统[15-16],为了匹配整个系统,需要调整相机使其拥有公共视场,从而实现较大的表面形变测量。尽管多相机系统可以提高3D测量的范围和精度,但是由于测量视场受限、相机价格昂贵等因素的影响,使得多相机系统难以广泛应用于3D全场测量中。近年来,在外部设备的辅助下,DIC测量系统更易于实现3D全场测量。Yu等[17]采用单相机与四平面镜结合用于3D形变测量,镜面辅助的伪立体成像系统由两个虚拟相机组成,等价于从两个视角观察被测件,并将被测件的表面图像信息记录到摄像机传感器的两半上。Ge等[18]利用单相机和六平面镜结合实现3D全景测量,通过光路转换将单相机拓展为四台虚拟相机。Wang等[19]利用单彩色相机与分束器结合,通过提取红绿蓝(RGB)图像中三通道信息最终实现复杂试件轮廓的精确测量。以上方法均利用单相机构成两个或多个伪相机,降低了实验成本。随后,Chen等[20-22]利用双相机与双平面镜结合,通过在镜面上喷涂散斑或者改变双镜面之间的夹角获得镜面的反射变换矩阵,最终实现3D重建和形变测量。为了使操作更加简便,Li等[23]同样利用基于相机标定的坐标转换关系,最终实现全场形变测量。还有学者将棱镜或投影仪辅助相机来实现3D测量[24-26],相比于平面镜辅助的DIC测量系统,其测量视场有限。
鉴于现有单相机和多相机系统存在的缺陷,本文提出了一种双平面镜辅助的双相机3D重建方法,解决了单相机系统测量精度不高以及多相机系统标定复杂、成本昂贵等问题。为了提高镜面辅助DIC测量系统的测量精度,本文利用两异面射线公垂线中点[27]进行目标点3D重构,通过求解与放置在镜面前的棋盘格标定板角点有对应关系的镜面位置方程,最终实现被测件表面的3D重建,避免了在镜面上喷涂散斑占用相机的空间分辨率,使反射变换矩阵的求解变得简便且高效,解决了反射变换矩阵求解复杂且测量精度不高的问题。该方法所需的镜面尺寸更小,反射镜之间的夹角可调整的范围更大,且实验装置简单、测量精度高,在获取被测件表面信息及3D重建方面具有显著优势。
2 双平面镜双相机DIC测量系统
2.1 测量原理
对双平面镜双相机DIC测量系统进行完整搭建,如
2.2 3D坐标点匹配过程
从左相机拍摄得到的参考图像中的每个感兴趣表面分别指定一个感兴趣区域(ROI),用虚线框区域来划分不同表面部分,如
图 2. 基于子集的双目DIC测量系统示意图
Fig. 2. Schematic diagram of subset-based binocular DIC measurement system
以左相机参考图像的ROI1为例,首先在ROI1中指定计算点
式中:
2.3 虚实像反射变换矩阵
物体表面经过平面镜反射为虚像,通过反射变换将虚像转换至其真实位置,如
同时可得
由
将
将
式中,
因
2.4 成像畸变校正
对于普通的工业相机,二阶畸变校正即可满足对原始图像的校正,考虑到相机制造精度及组装精度,为了实现双相机的高精度标定,采用三阶径向畸变系数进行畸变矫正,其校正模型表示为
切向畸变校正模型表示为
式中:
双相机DIC测量系统对标定精度要求较高,为了提高系统测量精度,当完成对相机内外参数和畸变系数的初步估计后,对得到的初步标定结果进行优化。捆绑调整作为一种应用广泛的优化模型,结合收敛速度快、稳定性强的Levenberg-Marquardt算法可以对前期得到的全局参数进一步优化来提高标定精度,其目标函数表示为
式中:
2.5 平面镜位置标定
为了实现全场测量,需要标定两镜面的位置方程,以确定反射变换矩阵实现虚实像的转换。如
双目立体视觉模型如
图 5. 基于公垂线中点法进行角点的3D重构
Fig. 5. 3D reconstruction of corner points based on common vertical midpoint method
设任一棋盘格角点
式中,
两射线的方向向量分别表示为
式中:
令公垂线与两射线交点分别表示为
可求得
将
如
式中,
因棋盘格标定板位置改变一次便可求得一组新的镜面3D点,通过多次移动标定板并采用下式对
式中,
3 实验与结果
为了验证所提方法的准确性,分别对被测件进行静态实验和动态实验。在静态实验中,对游戏币的轮廓进行3D重建;在动态实验中,对五棱铝柱热变形进行研究。
3.1 整体实验装置
镜面辅助双目DIC测量系统如
图 6. 实验装置图。(a)镜面辅助双目DIC测量系统;(b)硬币实物图;(c)五棱铝柱实物图;(d)加热炉
Fig. 6. Diagram of experimental setup. (a) Mirror-assisted binocular DIC measurement system; (b) physical picture of coin; (c) physical picture of five-edge aluminum column; (d) heating furnace
3.2 静态实验
3.2.1 实验过程
在静态实验中,按
图 7. 部分镜面位置标定图像。(a)左相机;(b)右相机
Fig. 7. Partial calibration images of mirror position. (a) Left camera; (b) right camera
3.2.2 游戏币3D重建结果
通过所提方法,可以得到不同夹角时的两镜面方程及反射变换矩阵,镜面夹角调整为108°时求得的结果如
表 1. 两镜面方程及反射变换矩阵
Table 1. Two mirror equations and reflection transformation matrix
|
对普通游戏币的正反轮廓进行3D重建,游戏币的理论厚度为1.80 mm,测量厚度在1.75 mm左右,实物图如
图 9. 游戏币3D重建。(a)本文所提方法重建效果;(b)镜面喷涂散斑方法重建效果
Fig. 9. 3D reconstruction of game coins. (a) Reconstruction effect of proposed method; (b) reconstruction effect of method of spraying diffuse spots on mirror surface
鉴于在镜面上喷涂散斑的方法被广泛应用和研究,为了验证所提方法的3D重建精度,将其与在镜面上喷涂散斑的重建方法进行对比[22],镜面喷斑方法对游戏币3D重建效果如
3.3 动态实验
3.3.1 实验过程
在动态实验中,同样按
3.3.2 铝柱3D重建及热变形测量结果
通过所提方法得到反射变换矩阵,对三部分点云进行虚实转换并统一到世界坐标系统中,得到五棱铝柱表面轮廓的重建结果,如
试件温度为330 ℃时采集的图像为参考图像,20 ℃时采集的图像为目标图像,提取参考图像的所有轮廓点,选取试件
图 11. 铝柱五个外侧表面的热形变结果。(a) A面;(b) B面;(c) C面;(d) D面;(e) E面
Fig. 11. Thermal deformation results of five outer surfaces of aluminum column. (a) Surface A; (b) surface B; (c) surface C; (d) surface D; (e) surface E
表 2. 五棱铝柱起止温度下不同表面的高度
Table 2. Heights of different surfaces at starting and ending temperatures of aluminum column
|
为了直观地对比降温过程中不同表面的高度变化,对于采集的1050帧图像每50帧作为一个节点,进行对比分析可得,五棱铝柱的外侧表面在降温过程中沿高度方向的变化曲线几乎一致,如
图 12. 降温过程A~E面的高度变化趋势
Fig. 12. Height variation trend of plane A to E during temperature reduction
图 13. 降温过程中两种方法的高度变化偏差值
Fig. 13. Deviation values of height changes during temperature reduction for both methods
3.3.3 热变形仿真结果
作为对比,采用Ansys仿真软件分析五棱柱体的热变形情况,零件的物理参数如
表 3. 铝的物理参数
Table 3. Physical parameters of aluminum
|
图 14. 采用Ansys对试件进行网格划分及热变形仿真。(a)有限元网格划分图;(b)热变形仿真结果
Fig. 14. Mesh division and thermal deformation simulation of specimen by Ansys. (a) Finite element meshing map; (b) simulation result of thermal deformation
4 结论
本文提出了一种双平面镜辅助的双相机视觉测量的方法,通过平面镜的光路转换实现从不同角度对被测件进行3D测量。静态实验结果表明,所提方法对游戏币的3D重建效果优于在镜面上喷涂散斑的重建方法,且精度较高。动态实验结果表明,当五棱铝柱温度从330 ℃降至20 ℃时,铝柱外侧表面高度变化与有限元软件仿真结果基本一致,且所提方法测量得到的高度变化偏差值均小于镜面喷涂散斑的方法。可见,所提方法避免了在镜面上喷涂散斑占用相机的空间分辨率,且操作简单,测量精度高,有良好的应用前景。
[1] 潘济宇, 张水强, 苏志龙, 等. 基于数字图像相关的水下螺旋桨三维变形测量[J]. 光学学报, 2021, 41(12): 1212001.
[2] 韩伟, 吴丹. 光学显微镜图像恢复及在数字图像相关方法应变测量中的应用[J]. 激光与光电子学进展, 2023, 60(14): 1410004.
[3] Hu P Y, Yang S M, Zheng F H, et al. Accurate and dynamic 3D shape measurement with digital image correlation-assisted phase shifting[J]. Measurement Science and Technology, 2021, 32(7): 075204.
[4] Srivastava V, Baqersad J. An optical-based technique to obtain operating deflection shapes of structures with complex geometries[J]. Mechanical Systems and Signal Processing, 2019, 128: 69-81.
[6] Li Y R, Zhao D, Ma X Y, et al. Panoramic digital image correlation for 360-deg full-field displacement measurement[J]. Applied Sciences, 2023, 13(3): 2019.
[7] Xia S, Gdoutou A, Ravichandran G. Diffraction assisted image correlation: a novel method for measuring three-dimensional deformation using two-dimensional digital image correlation[J]. Experimental Mechanics, 2013, 53(5): 755-765.
[8] Wu R, Zhang D S, Yu Q F, et al. Health monitoring of wind turbine blades in operation using three-dimensional digital image correlation[J]. Mechanical Systems and Signal Processing, 2019, 130: 470-483.
[9] Ye M T, Liang J, Li L G, et al. Simultaneous measurement of external and internal surface shape and deformation based on photogrammetry and stereo-DIC[J]. Optics and Lasers in Engineering, 2022, 158: 107179.
[10] Xie R L, Chen B, Pan B. Mirror-assisted multi-view high-speed digital image correlation for dual-surface dynamic deformation measurement[J]. Science China Technological Sciences, 2023, 66(3): 807-820.
[11] Pankow M, Justusson B, Waas A M. Three-dimensional digital image correlation technique using single high-speed camera for measuring large out-of-plane displacements at high framing rates[J]. Applied Optics, 2010, 49(17): 3418-3427.
[12] Chen Y X, Wang B, Li Q W, et al. Field-of-view-enlarged single-camera 3-D shape reconstruction[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 5009212.
[13] Orteu J J, Bugarin F, Harvent J, et al. Multiple-camera instrumentation of a single point incremental forming process pilot for shape and 3D displacement measurements: methodology and results[J]. Experimental Mechanics, 2011, 51(4): 625-639.
[14] Li J R, Xie X, Yang G B, et al. Whole-field thickness strain measurement using multiple camera digital image correlation system[J]. Optics and Lasers in Engineering, 2017, 90: 19-25.
[15] Chen F X, Chen X, Xie X, et al. Full-field 3D measurement using multi-camera digital image correlation system[J]. Optics and Lasers in Engineering, 2013, 51(9): 1044-1052.
[16] Solav D, Moerman K M, Jaeger A M, et al. A framework for measuring the time-varying shape and full-field deformation of residual limbs using 3-D digital image correlation[J]. IEEE Transactions on Bio-Medical Engineering, 2019, 66(10): 2740-2752.
[17] Yu L P, Pan B. Single-camera stereo-digital image correlation with a four-mirror adapter: optimized design and validation[J]. Optics and Lasers in Engineering, 2016, 87: 120-128.
[18] 葛朋祥, 王欢庆, 朱奕磊, 等. 基于平面镜成像的单相机数字图像相关三维全景测量[J]. 中国激光, 2022, 49(9): 0904004.
[19] Wang Y H, Dan X Z, Li J R, et al. Multi-perspective digital image correlation method using a single color camera[J]. Science China Technological Sciences, 2018, 61(1): 61-67.
[20] Chen B, Zhao J, Pan B. Mirror-assisted multi-view digital image correlation with improved spatial resolution[J]. Experimental Mechanics, 2020, 60(3): 283-293.
[21] Chen B, Genovese K, Pan B. In vivo panoramic human skin shape and deformation measurement using mirror-assisted multi-view digital image correlation[J]. Journal of the Mechanical Behavior of Biomedical Materials, 2020, 110: 103936.
[22] Chen B, Pan B. Mirror-assisted panoramic-digital image correlation for full-surface 360-deg deformation measurement[J]. Measurement, 2019, 132: 350-358.
[23] 李桂华, 李涛, 孙卫庆, 等. 基于双平面镜双相机组合测量全场变形的方法[J]. 光学学报, 2023, 43(2): 0212007.
[24] Li Y, Fu Y J, Zhong K J, et al. A virtual binocular line-structured light measurement method based on a plane mirror[J]. Optics Communications, 2022, 510: 127974.
[25] Shao X X, Qu J Y, Chen W W. Single-camera three-dimensional digital image correlation with enhanced accuracy based on four-view imaging[J]. Materials, 2023, 16(7): 2726.
[26] Wu L F, Zhu J G, Xie H M. Single-lens 3D digital image correlation system based on a bilateral telecentric lens and a bi-prism: validation and application[J]. Applied Optics, 2015, 54(26): 7842-7850.
[27] 徐巧玉, 姚怀, 齐鑫, 等. 基于公垂线约束的立体视觉建模方法[J]. 仪器仪表学报, 2008, 29(11): 2430-2434.
Xu Q Y, Yao H, Qi X, et al. Stereo vision reconstruction method based on common perpendicular constraint[J]. Chinese Journal of Scientific Instrument, 2008, 29(11): 2430-2434.
[28] 李世林, 戴松新, 胡中文, 等. 基于亚像素图像拼接的透明元件视觉测量方法[J]. 激光与光电子学进展, 2023, 60(6): 0615004.
[29] Su Y, Zhang Q C, Xu X H, et al. Interpolation bias for the inverse compositional Gauss-Newton algorithm in digital image correlation[J]. Optics and Lasers in Engineering, 2018, 100: 267-278.
Article Outline
李桂华, 王紫威, 孙卫庆, 葛朋祥, 王浩宇, 张梅. 双平面镜双相机数字图像相关技术在三维重建中的应用[J]. 光学学报, 2024, 44(8): 0812003. Guihua Li, Ziwei Wang, Weiqing Sun, Pengxiang Ge, Haoyu Wang, Mei Zhang. Application of Dual Plane-Mirror Dual-Camera Digital Image Correlation Technology in Three-Dimensional Reconstruction[J]. Acta Optica Sinica, 2024, 44(8): 0812003.