WGAN-based surface polarization detection of ring forgings
-
摘要: 为了增强对环形锻件表面的质量检测能力，提出一种基于DoFP偏振相机的FSV偏振检测方法，并利用改进的WGAN生成对抗网络生成S参量图，将理论上必须通过两次采集才能实现的FSV偏振检测成功减少到只需一次采集。通过搭建偏振检测实验平台，建立S参量偏振图像数据集，训练WGAN神经网络。利用实验生成的偏振数据，得到增强的二维表面图像，并进行相应的偏振三维建模。所提出的方法可同时实现对LCVR的快轴角度和相位延迟量的同步自标定，极大地便利了工业上的应用。实验结果表明，与传统成像系统和非FSV偏振成像系统相比，所提出的检测方法测得的DoP图像在图像评价指标上提升3%以上，大大提高了对环形锻件表面的质量检测能力。
-
关键词:
- 环形锻件 /
- WGAN生成对抗神经网络 /
- 偏振成像 /
- 同步自标定 /
- 全斯托克斯矢量(FSV)
Abstract: A DoFP polarization camera-based FSV polarization detection method is proposed in order to enhance the quality inspection of annular forging surfaces. An improved WGAN generating adversarial network is utilized to generate S-parametric maps to reduce the successful FSV polarization detection, which theoretically must be achieved by two acquisitions, to merely one acquisition. By constructing an experimental platform for polarization detection, an S-parameter polarization image dataset is established and the WGAN neural network is trained. Through the experimentally generated polarization data, an enhanced 2D surface image is obtained and the corresponding 3D modeling of polarization is performed. The proposed method enables simultaneous self-calibration of the fast-axis angular and phase delay quantities of LCVR, which greatly facilitates industrial applications. The experimental results present that the DoP images measured by the proposed detection method are improved by more than 3% in image evaluation index compared with the conventional imaging system and non-FSV polarization imaging system, which greatly improves the quality inspection capability of the ring forging surface. -
表 1 EWV全局优化数值表
$ \delta $/(°) $ \beta $/(°) $ {\delta }_{1} $/(°) $ {\delta }_{2} $/(°) $ {V}_{ewv} $ $ \delta \in \left[\mathrm{0,360}\right] $ 70.836 270.215 161.132 5.500$ {\sigma }^{2} $ $ \delta \in \left[\mathrm{0,180}\right] $ 72.152 90.001 29.721 5.500$ {\sigma }^{2} $ $ {\delta }_{1}=0, $
$ \delta \in \left[\mathrm{0,360}\right] $122.857 0 269.964 5.500$ {\sigma }^{2} $ $ {\delta }_{1}=0, $
$ \delta \in \left[\mathrm{0,180}\right] $96.505 0 90.001 5.500$ {\sigma }^{2} $ 
下载: 导出CSV
表 2 偏振图像评价指标
评价指标 偏振图像 $ I\left({S}_{0}\right) $ $ \mathrm{A}\mathrm{o}\mathrm{P} $ $ \mathrm{D}\mathrm{o}\mathrm{l}\mathrm{P} $ $ \mathrm{D}\mathrm{o}\mathrm{P} $ H 6.449 2 7.349 6 6.701 8 6.775 4 F 147.552 0 96.517 7 167.652 6 174.195 6 $ {M}_{\mathrm{M}\mathrm{S}\mathrm{E}} $ 16.298 4 19.490 5 17.453 2 18.591 9 $ {R}_{\mathrm{P}\mathrm{S}\mathrm{N}\mathrm{R}} $ 3.127 0 6.908 1 3.242 4 3.419 8 $ {M}_{\mathrm{S}\mathrm{S}\mathrm{I}\mathrm{M}} $ 0.004 2 0.003 9 0.009 1 0.010 0
下载: 导出CSV
-
[1] Bhattacharya S, Risi R D, Lombardi D, et al. On the seismic analysis and design of offshore wind turbines[J]. Soil Dynamics and Earthquake Engineering, 2021, 145: 106692. doi: 10.1016/j.soildyn.2021.106692 [2] 朱玉龙, 赵迎松, 方阳, 等. 孔边裂纹的旋转涡流检测技术研究[J/OL]. 中国机械工程: 1-11[2022-06-02]. http://kns.cnki.net/kcms/detail/42.1294.TH.20220322.1713.008.html [3] 杨传礼, 张修庆. 基于机器视觉和深度学习的材料缺陷检测应用综述[J/OL]. 材料导报, 2022(16): 1-19[2022-06-02]. http://kns.cnki.net/kcms/detail/50.1078.TB.20210728.1056.009.html [4] 周强国, 黄志明, 周炜. 偏振成像技术的研究进展及应用[J]. 红外技术, 2021, 43(9): 817-828. [5] 黎海育, 李抄, 李校博, 等. 基于偏振相机的全斯托克斯偏振仪优化研究[J]. 光学学报, 2020, 40(3): 167-174. [6] 李校博. Stokes矢量及Mueller矩阵的优化测量方法研究[D]. 天津: 天津大学, 2020. [7] Vedel M, Breugnot S, Lechocinski N. Full stokes polarization imaging camera[J]. International Society for Optics and Photonics, 2011, 8160: 892491. [8] Wei Y, Han P, Liu F, et al. Enhancement of underwater vision by fully exploiting the polarization information from the stokes vector[J]. Optics Express, 2021, 29(14): 22275-22287. doi: 10.1364/OE.433072 [9] 白杨, 赵开春, 尤政. 分焦平面偏振图像传感器偏振主轴方向的标定[J]. 光学精密工程, 2022, 30(1): 31-37. doi: 10.37188/OPE.20223001.0031 [10] Peinado A, Lizana A, Iemmi C, et al. Polarization imaging with enhanced spatial resolution[J]. Optics Communications, 2015, 338: 95-100. doi: 10.1016/j.optcom.2014.09.079 -
下载:
点击查看大图
图(10) / 表(2)
计量
- 文章访问数: 51
- HTML全文浏览量: 10
- PDF下载量: 19
- 被引次数: 0
