结合灰度信息的压敏漆图像配准

梁诚, 蒲方圆, 梁磊, 等. 结合灰度信息的压敏漆图像配准[J]. 光电工程, 2019, 46(2): 180301. doi: 10.12086/oee.2019.180301
引用本文: 梁诚, 蒲方圆, 梁磊, 等. 结合灰度信息的压敏漆图像配准[J]. 光电工程, 2019, 46(2): 180301. doi: 10.12086/oee.2019.180301
Liang Cheng, Pu Fangyuan, Liang Lei, et al. Pressure sensitive paint image registration combined with gray level information[J]. Opto-Electronic Engineering, 2019, 46(2): 180301. doi: 10.12086/oee.2019.180301
Citation: Liang Cheng, Pu Fangyuan, Liang Lei, et al. Pressure sensitive paint image registration combined with gray level information[J]. Opto-Electronic Engineering, 2019, 46(2): 180301. doi: 10.12086/oee.2019.180301

结合灰度信息的压敏漆图像配准

  • 基金项目:
    教育部春晖计划(z2016149);西华大学重点实验室项目(szjj2017-065)
详细信息
    作者简介:
    通讯作者: 梁磊(1976-),男,高级工程师,主要从事风洞测试技术方面的研究。E-mail: skywork@163.com
  • 中图分类号: V211.753;TP751.1

Pressure sensitive paint image registration combined with gray level information

  • Fund Project: Supported by the Ministry of Education Chunhui Project (z2016149) and Xihua University Key Laboratory Development Program (szjj2017-065)
More Information
  • 压敏漆技术是一种经济性高、速度快的风洞测压前沿技术。在风洞试验中,由于强风影响,模型会发生畸变,造成有风图像和无风图像难以配准,从而严重影响测压精度。针对这一问题,本文创新性的将二维非刚性ICP算法用于此问题,采用点云方式使得图像细节区域有效配准,同时也有利于后续三维重建工作。然而由于二维非刚性ICP算法仅考虑二维坐标位置关系,忽略压敏漆图像像素灰度具有的相关性,使得配准精度不高。直接利用三维非刚性ICP算法又会发生误配准,所以为了进一步提高配准精度,本文提出了一种基于像素关联搜索策略的非刚性ICP算法,算法设计了综合考虑2D坐标与像素灰度值的双目标搜索策略,实现了精确的局部匹配点搜索与双目标优化。在多组压敏漆图像上将本文算法与五种配准算法进行了对比实验分析。实验结果表明,本文所提出的算法具有最好的配准精度。相比次优算法,RMSE提升超过15%,NMI提升在5%左右。

  • Overview: In recent years, with the rapid development of China′s aviation industry, higher requirements have been put forward for designing and manufacturing new types of aircraft. The wind tunnel pressure test is a key link in design and manufacturing. The current popular method is to use pressure-sensitive paint technology with the advantages of economy and speed for pressure measurement. The principle is that the pressure-sensitive paint will perform a dynamic oxidation quenching reaction according to the surface pressure of the test model, thus emitting light with different intensities. The two images of the test model under windy and windless conditions are taken and calculated using the Stern-Volmer formula. The corresponding pixel pressure data is obtained, and the two-dimensional pressure distribution of the entire model surface can be obtained. However, due to the existence of aerodynamic loads in the actual test, deformation and displacement of the test model will occur. In particular, under strong wind speed conditions, the test model will easily undergo elastic deformation and the resulting image will also be distorted. Therefore, registration of windy images is required. However, current algorithms for registering pressure-sensitive paint images have more or less partial disadvantages. For example, it is hard to achieve registration of image details for the registration of the model using only a rigid transformation, or simple non-rigid transformation registration of the model as a whole (affine transformation or polynomial transformation). This paper innovatively uses the two-dimensional non-rigid iterative closest point (ICP) algorithm to solve those problems. The point cloud method is used to make the image detail area to be effectively registered, and it is also conducive to the subsequent three-dimensional reconstruction work. However, due to the traditional two-dimensional non-rigid ICP algorithm, only the two-dimensional coordinate position relationship is considered, and the correlation of pixel grayscales of the pressure-sensitive paint image is ignored. Thus, the registration accuracy is not too high. However, if the three-dimensional non-rigid ICP algorithm is directly used, misregistration will occur. Therefore, in order to further improve the accuracy of registration, this paper proposes an improved non-rigid ICP algorithm based on pixel association search strategy. The algorithm designs a dual-target search strategy that takes 2D coordinates and pixel gray values into consideration and achieves accurate local matching, realizing point search and double goal optimization. The algorithm of this paper is compared with five registration algorithms on multiple sets of pressure sensitive paint images. The experimental results show that the proposed algorithm has the best registration accuracy. Compared to the suboptimal algorithm, RMSE increased by more than 15%, and NMI increased by about 5%.

  • 加载中
  • 图 1  畸变图像(a)与原图(b)对比图

    Figure 1.  Comparison between distortion image and original image

    图 2  局部网格节点图

    Figure 2.  Local grid node diagram

    图 3  x-pixel intensity三点截面图

    Figure 3.  x-pixel intensity three-point longitudinal section

    图 4  x-pixel intensity多点截面图

    Figure 4.  x-pixel intensity multi-point longitudinal section

    图 5  点云提取图像。(a)活动轮廓模型提取轮廓;(b)所得模型二维网格图像

    Figure 5.  Point cloud extraction image. (a) Extracted contour by active contour model; (b) 2D grid image of the model

    图 6  0off/10各算法配准后重叠图像。(a)原始重叠图像;(b) MSID配准后重叠图像;(c) MI-Bspline配准后重叠图像;(d) NRICP配准后重叠图像;(e) NRICP(CnNMI)配准后重叠图像;(f) NRICP(FnNMI)配准后重叠图像;(g)本文算法

    Figure 6.  Overlapped images after registration of different algorithms in the case of 0off/10. (a) Raw overlapped images; (b) Overlapped images after MSID registration; (c) Overlapped images after MI-Bspline registration; (d) Overlapped images after NRICP registration; (e) Overlapped images after NRICP (CnNMI) registration; (f) Overlapped images after NRICP (FnNMI) registration; (g) Overlapped images based on our algorithm

    图 7  20off/20D各算法配准后重叠图像。(a)原始重叠图像;(b) MSID配准后重叠图像;(c) MI-Bspline配准后重叠图像;(d) NRICP配准后重叠图像;(e) NRICP(CnNMI)配准后重叠图像;(f) NRICP(FnNMI)配准后重叠图像;(g)本文算法

    Figure 7.  Overlapped images after registration of different algorithms in the case of 20off/20D. (a) Raw overlapped images; (b) Overlapped images after MSID registration; (c) Overlapped images after MI-Bspline registration; (d) Overlapped images after NRICP registration; (e) Overlapped images after NRICP (CnNMI) registration; (f) Overlapped images after NRICP (FnNMI) registration; (g) Overlapped images based on our algorithm

    表 1  0off/10各算法配准后指标对比表

    Table 1.  Comparison of indicators after registration of different algorithms in the case of 0off/10

    Raw (a) MSID (b) MI-Bspline (c) NRICP (d) NRICP(CnNMI) (e) NRICP(FnNMI) (f) Ours (g)
    RMSE 3.8194 3.0133 1.4077 2.7612 1.5211 1.215 1.0529
    NMI 0.4721 0.5564 0.5710 0.4892 0.5587 0.5658 0.5970
    下载: 导出CSV

    表 2  20off/20D各算法配准后指标对比

    Table 2.  Comparison of indicators after registration of different algorithms in the case of 20off/20D

    Raw (a) MSID (b) MI-Bspline (c) NRICP (d) NRICP(CnNMI) (e) NRICP(FnNMI) (f) Ours (g)
    RMSE 2.8611 1.8630 1.3591 1.3482 1.2808 1.2460 1.0065
    NMI 0.5508 0.5891 0.6156 0.5897 0.5926 0.5954 0.6212
    下载: 导出CSV
  • [1]

    Friedl F, Krah N, Jähne B. Optical sensing of oxygen using a modified stern-volmer equation for high laser irradiance[J]. Sensors and Actuators B: Chemical, 2015, 206: 336-342. doi: 10.1016/j.snb.2014.09.073

    [2]

    Peng D, Chen J W, Jiao L R, et al. A fast-responding semi-transparent pressure-sensitive paint based on through-hole anodized Aluminum oxide membrane[J]. Sensors and Actuators A: Physical, 2018, 274: 10-18. http://cn.bing.com/academic/profile?id=eda48a999a927572fe14e800ef5cf39e&encoded=0&v=paper_preview&mkt=zh-cn

    [3]

    Kontis K. A review of some current research on pressure sensitive paint and thermographic phosphor techniques[J]. The Aeronautical Journal, 2007, 111(1122): 495-508. doi: 10.1017/S0001924000004747

    [4]

    Park S H, Sung H J. Correlation-based image registration for applications using pressure-sensitive paint[J]. AIAA Journal, 2005, 43(3): 472-478. doi: 10.2514/1.7661

    [5]

    Fujimatsu N, Tamura Y, Fujii K. Improvement of noise filtering and image registration methods for the pressure sensitive paint experiments[J]. Journal of Visualization, 2005, 8(3): 225-233. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=04555f5a71ef13d144d24b7f0fa86c16

    [6]

    陈畅, 陈勇, 黄琦.压敏漆技术中的图像几何畸变校正研究[J].计算机工程与设计, 2010, 31(24):5298-5301. http://d.old.wanfangdata.com.cn/Periodical/jsjgcysj201024035

    Chen C, Chen Y, Huang Q. Study of image geometric distortion correction for PSP measurements[J]. Computer Engineering and Design, 2010, 31(24): 5298-5301. http://d.old.wanfangdata.com.cn/Periodical/jsjgcysj201024035

    [7]

    高丽敏, 韦楠, 高杰, 等.基于内流场PSP测量技术的图像后处理[J].实验流体力学, 2013, 27(1): 93-97. doi: 10.3969/j.issn.1672-9897.2013.01.017

    Gao L M, Wei N, Gao J, et al. Image processing of PSP technique in the internal flow[J]. Journal of Experiments in Fluid Mechanics, 2013, 27(1): 93-97. doi: 10.3969/j.issn.1672-9897.2013.01.017

    [8]

    熊红亮, 向星居, 郎卫东.压敏漆图像数据处理技术[J].实验流体力学, 2011, 25(2): 77-82. doi: 10.3969/j.issn.1672-9897.2011.02.016

    Xiong H L, Xiang X J, Lang W D. Image data processing techniques for pressure sensitive paint[J]. Journal of Experiments in Fluid Mechanics, 2011, 25(2): 77-82. doi: 10.3969/j.issn.1672-9897.2011.02.016

    [9]

    Lu X S, Zhang S, Yang W, et al. SIFT and shape information incorporated into fluid model for non-rigid registration of ultrasound images[J]. Computer Methods and Programs in Biomedicine, 2010, 100(2): 123-131. doi: 10.1016/j.cmpb.2010.03.005

    [10]

    Aganj I, Iglesias E I, Reuter M, et al. Mid-space-independent deformable image registration[J]. Neuroimage, 2017, 152: 158-170. doi: 10.1016/j.neuroimage.2017.02.055

    [11]

    Besl P J, McKay N D. A method for registration of 3-D shapes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(2): 239-256. doi: 10.1109/34.121791

    [12]

    Chi Y, Yu X Q, Luo Z Y. 3D point cloud matching based on principal component analysis and iterative closest point algorithm[C]//Proceedings of 2016 International Conference on Audio, Language and Image Processing, Shanghai, China, 2016: 404-408.

    [13]

    Amberg B, Romdhani S, Vetter T. Optimal step nonrigid ICP algorithms for surface registration[C]//Proceedings of 2007 IEEE Conference on Computer Vision and Pattern Recognition, Minneapolis, MN, USA, 2007: 1-8.

    [14]

    Hontani H, Matsuno T, Sawada Y. Robust nonrigid ICP using Outlier-Sparsity regularization[C]//Proceedings of 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI, USA, 2012: 174-181.

    [15]

    Cheng S Y, Marras I, Zafeiriou S, et al. Statistical non-rigid ICP algorithm and its application to 3D face alignment[J]. Image and Vision Computing, 2017, 58: 3-12. doi: 10.1016/j.imavis.2016.10.007

    [16]

    Chen J, Ma J Y, Yang C C, et al. Non-rigid point set registration via coherent spatial mapping[J]. Signal Processing, 2015, 106: 62-72. doi: 10.1016/j.sigpro.2014.07.004

    [17]

    Lu M, Zhao J, Guo Y L, et al. Accelerated coherent point drift for automatic Three-Dimensional point cloud registration[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(2): 162-166. doi: 10.1109/LGRS.2015.2504268

    [18]

    Rueckert D, Sonoda L I, Hayes C, et al. Nonrigid registration using free-form deformations: application to breast MR images[J]. IEEE Transactions on Medical Imaging, 1999, 18(8): 712-721. doi: 10.1109/42.796284

    [19]

    Woo J, Stone M, Prince J L. Multimodal registration via mutual information incorporating geometric and spatial context[J]. IEEE Transactions on Image Processing, 2015, 24(2): 757-769. http://cn.bing.com/academic/profile?id=d26b4ce110d299e272ae0b2248f361cb&encoded=0&v=paper_preview&mkt=zh-cn

    [20]

    Qiu Z Y, Tang H H, Tian D S. Non-rigid medical image registration based on the thin-plate spline algorithm[C]//Proceedings of 2009 WRI World Congress on Computer Science and Information Engineering, Los Angeles, CA, USA, 2009: 522-527.

    [21]

    Chan T F, Sandberg B Y, Vese L A. Active contours without edges for vector-valued images[J]. Journal of Visual Communication and Image Representation, 2000, 11(2): 130-141. doi: 10.1006/jvci.1999.0442

    [22]

    Xu L, Wan J W L, Bian T T. A continuous method for reducing interpolation artifacts in mutual Information-Based rigid image registration[J]. IEEE Transactions on Image Processing, 2013, 22(8): 2995-3007. doi: 10.1109/TIP.2013.2251644

    [23]

    Collignon A. Automated multimodality image registration using information theory[C]//Proceedings of International Conference Information Processing in Medical Imaging, Ile De Berder, France, 1995: 263-274.

    [24]

    Viola P, Wells III W M. Alignment by maximization of mutual information[J]. International Journal of Computer Vision, 1997, 24(2): 137-154. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=ebc82ea41912813953d4150e177b31be

    [25]

    Studholme C, Hill D L G, Hawkes D J. An overlap invariant entropy measure of 3D medical image alignment[J]. Pattern Recognition, 1999, 32(1): 71-86. http://cn.bing.com/academic/profile?id=e3a028b73d05b4f14ee07db7fe29a482&encoded=0&v=paper_preview&mkt=zh-cn

    [26]

    Friedman J H, Bentley J L, Finkel R A. An algorithm for finding best matches in logarithmic expected time[J]. ACM Transactions on Mathematical Software, 1977, 3(3): 209-226. doi: 10.1145/355744.355745

  • 加载中

(7)

(2)

计量
  • 文章访问数:  7246
  • PDF下载数:  2449
  • 施引文献:  0
出版历程
收稿日期:  2018-06-01
修回日期:  2018-10-17
刊出日期:  2019-02-01

目录

/

返回文章
返回