-
实验平台采用650 nm的线激光发生器、Daheng MER2-503-36 U3 M的工业相机和型号GCD-0401 M、控制旋转精度为0.001 °旋转台控制器。实验处理器配置为英特尔Core i5-10210 U CPU@1.60 GHz、四核、内存8 G,软件开发平台为Visual Studio 2019,OpenCV 4.0.0和PCL1.11.0。在标定视觉模块后(标定结果见表1)[11-12],以哑光陶瓷标准球棒和金属轮毂为被测物分别进行三维测量。
Parameter name Calibration results
Camera internal parameters$ \left[ {\begin{array}{*{20}{c}} {{\text{3\;553}}{\text{.406\;333}}}&{\text{0}}&{{\text{1\;251}}{\text{.233\;369}}} \\ {\text{0}}&{{\text{3\;556}}{\text{.068\;465}}}&{{\text{1\;020}}{\text{.443\;882}}} \\ {\text{0}}&{\text{0}}&{\text{1}} \end{array}} \right] $ Distortion coefficient [0.107 913 1.088 657 0.000 477 0.000 105 −7.125 999] Optical plane equation $ - {\text{0}}{\text{.053\;813\;3}}{x_c} + {\text{0}}{\text{.829\;647}}{y_c} + {\text{0}}{\text{.555\;689}}{z_c} - 93.540\;8 = 0 $ Rotation axis equation $\dfrac{{{x_c} - {\text{4}}{\text{.388\;52}}}}{{{\text{0}}{\text{.998\;333}}}} = \dfrac{{{y_c} + {\text{56}}{\text{.862\;8}}}}{{{\text{ 0}}{\text{.0289\;22}}}} = \dfrac{{{z_c} - {\text{263}}{\text{.342}}}}{{{\text{0}}{\text{.0499\;468}}}}$ Table 1. Vision module calibration parameters
-
通过对哑光陶瓷标准球棒进行线激光旋转扫描来重建球棒表面点云数据,通过最小二乘法拟合球方程进而计算标准球棒A,B两球的球半径和球心距,10组重复测量结果见表2。
Test groups A
diameter/mmB
diameter/mmSphere center distance between
A & B/mm1 25.471 2 25.475 8 100.046 2 2 25.459 2 25.473 0 100.051 7 3 25.444 8 25.475 8 100.083 0 4 25.459 2 25.475 4 100.073 5 5 25.440 8 25.481 8 100.039 2 6 25.446 8 25.342 8 100.021 3 7 25.379 2 25.467 2 100.076 2 8 25.386 1 25.458 7 100.062 3 9 25.462 5 25.461 7 100.057 6 10 25.458 0 25.358 2 100.044 5 Standard value 25.411 0 25.413 6 100.001 1 Maximum deviation 0.060 2 0.068 2 0.081 9 Standard deviation 0.042 5 0.058 9 0.057 9 Table 2. Standard bat measurement results
在相同实验环境下10组的测量结果有一定的波动,但总体的测量系统精度误差在±0.06 mm内。误差产生的原因除了线结构光旋转扫描视觉系统的标定带来的误差,还因为标准球棒表面线结构光散射较多,投影条纹宽度不一,导致线结构光投影条纹中心线提取不准确,造成后续三维点云重建产生偏差。但综合以上测量系统精度评价实验结果的分析,可知文中测量方案的测量精度较高,满足测量精度要求。
-
被测物金属轮毂是个高80 mm,最宽直径277 mm的旋转体。金属轮毂中部围绕了一圈宽度约7 mm的焊缝区域。实验目标是测量轮毂焊缝区域的外轮廓形貌和不同扫描角度下的最大半径值。被测金属轮毂实物图和实验图见图7。
将被测轮毂放置于旋转台上,以单位旋转角度0.307°进行扫描,共采集1172张线结构光投影条纹图像,分别利用文中提出的基于缺失区域自适应灰度增强的光条纹中心线提取方法和灰度重心法提取中心线,通过视觉模块和旋转扫描中心轴的标定结果对轮毂外轮廓形貌进行测量,重建效果见图8。由图8可以清晰的看到文中提出的基于缺失区域自适应灰度增强的光条纹中心线提取算法有效地修复了光条纹缺失区域,重建点云数量增加了969个,有效提高了点云完整性。
在相同实验环境下,对同一轮毂进行5组外轮廓形貌测量重复性实验,计算不同旋转扫描角度下的轮毂外轮廓最大半径值。重复测量结果如图9所示,可以直观看到每组轮毂外轮廓最大半径随扫描角度的变化和分布趋势一致。
通过计算每个扫描角度下5组轮毂最大半径测量值的标准差σ,并用$ \delta =\sigma /\stackrel{-}{r} $来表示轮毂外轮廓最大半径测量的重复性误差,其中$ \stackrel{-}{r} $为多次测量的平均值,显然$ \delta $越小测量重复性越高。测量重复性误差结果见图10。
轮毂外轮廓不同扫描位置下的最大半径测量重复性误差均优于0.3%,最大值为128.019°处的重复性误差0.2743%。综上说明文中基于线结构光旋转扫描的三维视觉测量方案在轮毂外轮廓形貌测量应用中可重复性较高。
Research on 3-D vision measurement technology based on line structured light rotating scanning and laser stripe repair
doi: 10.3788/IRLA20210894
- Received Date: 2021-12-01
- Rev Recd Date: 2022-01-10
- Available Online: 2022-03-04
- Publish Date: 2022-02-28
-
Key words:
- 3-D vision measurement /
- line structured light /
- calibration
Abstract: Non-contact 3-D vision measurement is widely used in industrial manufacturing quality inspection. Aiming at the application scenario of industrial metal parts detection, a 3-D vision measurement scheme based on line structured light rotating scanning and laser stripe repair was proposed. Firstly, through the computer vision technology based on line structured light projection, the line structured light rotating scanning vision subsystem was designed, and the industrial camera, line structured light plane and rotating scanning central axis were calibrated; Then, aiming at the problem of missing data in the low gray area of the collected laser stripe image, a laser stripe center line extraction algorithm based on adaptive gray enhancement of the missing area was proposed, which effectively repaired the line structured light projection stripes of the tested parts; At the same time, using the line structured light 3-D vision measurement scheme proposed in this paper, the accuracy of the measurement system was evaluated by reconstructing the surface point cloud of the standard bat and calculating the diameter and spacing of the two balls. The accuracy of the measurement system was better than 0.06 mm; Finally, the shape of the outer contour of the metal hub was measured, and the maximum radius of the outer contour of the hub was calculated through the repeatability experiment. It is verified that the repeatability error is better than 0.03%. The experimental results show that this method can realize the 3-D measurement of industrial metal parts without damage, high efficiency and high precision, and make up for the defects of the contact 3-D measurement method.