搜索

x

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于卷积神经网络的高精度分块镜共相检测方法

赵伟瑞 王浩 张璐 赵跃进 褚春艳

引用本文:
Citation:

基于卷积神经网络的高精度分块镜共相检测方法

赵伟瑞, 王浩, 张璐, 赵跃进, 褚春艳

High-precision co-phase method for segments based on a convolutional neural network

Zhao Wei-Rui, Wang Hao, Zhang Lu, Zhao Yue-Jin, Chu Chun-Yan
PDF
HTML
导出引用
  • 为获得与单口径望远镜相当的空间分辨率, 使成像系统成像质量达到或接近衍射极限, 拼接主镜式望远镜的分块子镜应确保实现共相位拼接, 本文针对拼接主镜式望远镜高精度平移(piston)误差检测问题, 提出了一种基于卷积神经网络的高精度平移误差检测方法. 通过在成像系统的出瞳面上设置具有离散孔的光阑, 构建了对平移误差极为敏感的点扩散函数图像数据集, 根据此数据集的特点搭建了具有高性能的网络模型, 并测试得到网络的最佳检测范围. 仿真结果表明, 在略小于一个波长的捕获范围内, 单个网络能够准确地输出一个或多个分块子镜的平移误差; 应用于六子镜成像系统时, 平移误差检测精度达0.0013λ RMS (root mean square), 并且方法对残余倾斜(tip-tilt)误差、波前像差、CCD噪声、光源带宽具有良好的鲁棒性. 该方法简单快速, 可广泛应用于分块镜系统的平移误差检测.
    In order to achieve the resolution comparable to the resolution of a monolithic primary mirror telescope and make the imaging quality of the imaging system reach or approach to the diffraction limit, the submirrors of the segments telescope should ensure co-phase splicing. To solve the problem of phase error detection, a high-precision piston error detection method is proposed based on convolutional neural network (CNN). By setting a mask with a sparse multi-subpupil configuration on the exit pupil of the imaging system, a point spread function (PSF) image dataset that is extremely sensitive to the piston error is constructed. According to the characteristics of this dataset, a high-performance CNN model is built. And the best detection range of CNN is tested. The simulation results show that a single network can accurately output the piston error of one or more submirrors in the capture range slightly less than one wavelength. When the single network is applied to the six-submirror imaging system, the detection precision of the piston error reaches an RMS value of 0.0013λ (here, RMS stands for root mean square). And the method has good robustness to residual tip-tilt error, wavefront aberration, and CCD noise, light source bandwidth. The method is simple and fast, and can be widely used to detect the piston error of the segments.
      通信作者: 赵伟瑞, zwrei@bit.edu.cn ; 褚春艳, xyzccy@163.com
    • 基金项目: 国家自然科学基金(批准号: 11874086)资助的课题
      Corresponding author: Zhao Wei-Rui, zwrei@bit.edu.cn ; Chu Chun-Yan, xyzccy@163.com
    • Funds: Project supported by the National Natural Science Foundation of China (Grant No. 11874086).
    [1]

    Hege E K, Beckers J M, Strittmatter P A, McCarthy D W 1985 Appl. Opt. 24 2565Google Scholar

    [2]

    罗群, 黄林海, 顾乃庭, 李斐, 饶长辉 2012 物理学报 61 069501Google Scholar

    Luo Q, Huang L H, Gu N T, Li F, Rao C H 2012 Acta Phys. Sin. 61 069501Google Scholar

    [3]

    颜召军, 陈欣扬, 郑立新, 丁媛媛, 朱能鸿 2016 物理学报 65 199501Google Scholar

    Yan Z J, Chen X Y, Zheng L X, Ding Y Y, Zhu N H 2016 Acta Phys. Sin. 65 199501Google Scholar

    [4]

    常军, 张正慧, 王蕊瑞 2011 物理学报 60 034218Google Scholar

    Chang J, Zhang Zh H, Wang R R 2011 Acta Phys. Sin. 60 034218Google Scholar

    [5]

    Platt B C, Shack R 2001 J. Refract. Surg. 17 S573Google Scholar

    [6]

    Harvey J E, Rockwell R A 1988 Opt. Eng. 27 762Google Scholar

    [7]

    Baron F, Cassaing F, Blanca A, Laubier D 2003 Opt. Eng. 4852 663Google Scholar

    [8]

    Toni F 2010 Phys. Today. 63 26Google Scholar

    [9]

    Chanan G, Troy M, Dekens F, Michaels S, Nelson J, Mast T, Kirkman D 1998 Appl. Opt. 37 140Google Scholar

    [10]

    Shi F, Chanan G, Ohara C, Troy M, David C 2004 Appl. Opt. 43 4474Google Scholar

    [11]

    Li D, Xu S Y, Wang D, Yan D J 2019 Opt. Lett. 44 1170Google Scholar

    [12]

    Pizarro C, Arasa J, Laguarta F, Tomàs N 2002 Appl. Opt. 41 4562Google Scholar

    [13]

    Roddier F 1990 Appl. Opt. 29 1402Google Scholar

    [14]

    Dohlen K, Langlois M, Lanzoni P, Mazzanti S, Vigan A, Montoya L, Hernandez E, Reyes M, Surdej I, Yaitskova N 2006 Proc. SPIE 6267 626734Google Scholar

    [15]

    Esposito S, Pinna E, Puglisi A, Tozzi A, Stefanini P 2005 Opt. Lett. 30 2572Google Scholar

    [16]

    Pinna E, Esposito S, Puglisi A, Pieralli F, Myers R M, Busoni L, Tozzi A, Stefanini P 2006 Proc. SPIE 6267 62672YGoogle Scholar

    [17]

    Wang S S, Zhu Q D, Zhao W R, Li L, Cao G R 2009 Chin. Opt. Lett. 07 1007

    [18]

    Jiang A M, Wang S, Dong Z C, Xue J W, Wang J Y, Dai Y F 2018 Appl. Opt. 57 2736Google Scholar

    [19]

    Salinas-Luna J, Luna E, Salas L, Cruz-González I, Cornejo-Rodríguez A 2004 Opt. Express 12 3719Google Scholar

    [20]

    Vasishta G, Bootha A J, Colavitaa M M, Johnsona R L, Ligona E R, Moore J D, Palmera D L 2003 Proc. SPIE 4838 824Google Scholar

    [21]

    Booth J A, Adams M T, Ames G H, Fowler J R, Rakoczy J M 2000 Proc. SPIE 4003 176Google Scholar

    [22]

    易红伟, 李英才, 樊超, 王矫 2008 光子学报 37 1373

    Yi H W, Li Y C, Fan C, Wang J 2008 Acta Photon. Sin. 37 1373

    [23]

    Dailos G R, Lara D G, Juan T S, Jose R M 2018 Opt. Lett. 43 4264Google Scholar

    [24]

    Hui M, Li W Q, Liu M, Dong L Q, Kong L Q, Zhao Y J 2020 Appl. Opt. 59 771Google Scholar

    [25]

    Ma X F, Xie Z L, Ma H T, Xu Y J, Ren G, Liu Y 2019 Opt. Express 27 16058Google Scholar

    [26]

    Clampin M 2008 Proc. SPIE 41 254Google Scholar

    [27]

    Cheetham A C, Tuthill P G, Sivaramakrishnan A, Lloyd J P 2012 Opt. Express 20 29457Google Scholar

    [28]

    Cheetham A, Cvetojevic N, Norris B, Sivaramakrishnan A, Tuthill P 2014 Opt. Express 22 12924Google Scholar

  • 图 1  光阑子孔径分布

    Fig. 1.  Sub-pupils distribution of the diaphragm.

    图 2  不同平移误差对应的PSF图像样本 (a) p = 0; (b) p= 0.3λ

    Fig. 2.  PSF image samples corresponding to different piston errors: (a) p = 0; (b) p = 0.3λ.

    图 3  CNN网络结构

    Fig. 3.  The network structure of CNN.

    图 4  Mlpconv结构图

    Fig. 4.  Structure diagram of Mlpconv.

    图 5  不同检测范围时设定平移值对应的CNN预测误差分布图 (a) [–0.5λ, 0.5λ]; (b) [–0.49λ, 0.49λ]; (c) [–0.48λ, 0.48λ]

    Fig. 5.  The distribution of CNN prediction errors corresponding to the piston value in different detection ranges: (a) [–0.5λ, 0.5λ] ; (b) [–0.49λ, 0.49λ] ; (c) [–0.48λ, 0.48λ].

    图 6  六子镜成像系统 (a) 分块镜(六边形)和离散光阑孔(圆形); (b) PSF图像

    Fig. 6.  Six submirror imaging system: (a) Segments (hexagons) and sparse subpupils (circles); (b) PSF image.

    图 7  六子镜成像系统测试结果分布情况 (a) 子镜2; (b) 子镜3; (c) 子镜4; (d) 子镜5; (e) 子镜6

    Fig. 7.  Distributions of errors over all the testing results for six submirror system: (a) Submirror 2; (b) submirror 3; (c) submirror 4; (d) submirror 5; (e) submirror 6.

    表 1  不同检测范围时方法的检测精度

    Table 1.  Detection precision of the method in different detection ranges.

    检测范围/λ± 0.6± 0.5± 0.49± 0.48± 0.47± 0.46± 0.4± 0.35
    精度/λ RMS0.140.0130.00650.00140.00170.00170.00140.0018
    下载: 导出CSV

    表 2  六子镜成像系统数据集划分情况

    Table 2.  Division of six submirror imaging system data set.

    数据集占比/%样本形状标签形状
    训练集80(16000, 128, 128, 1)(16000, 5)
    验证集10(2000, 128, 128, 1)(2000, 5)
    测试集10(2000, 128, 128, 1)(2000, 5)
    下载: 导出CSV

    表 3  不同倾斜值时方法的检测精度

    Table 3.  Detection precision of the method in different tip-tilt values.

    倾斜值/λ RMS0.0040.040.10.20.30.4
    精度/λ RMS0.00150.00400.00950.0120.0230.041
    下载: 导出CSV

    表 4  不同信噪比值时方法的检测精度

    Table 4.  Dtection precision of the method in different signal-to-noise ratio.

    ${R_{{\text{SN}}}}$/dB504540353025
    精度/λ RMS0.00330.00370.00580.00910.0190.073
    下载: 导出CSV

    表 5  不同像差时方法的检测精度

    Table 5.  Dtection precision of the method in different aberrations.

    像差值/λ RMS0.010.020.030.040.050.06
    精度/λ RMS0.00470.00950.0140.0190.0240.032
    下载: 导出CSV

    表 6  不同带宽时方法的检测精度

    Table 6.  Dtection precision of the method with different spectral widths.

    带宽Δλ/nm10–610–410–21510
    精度/λ RMS0.00140.00150.00150.00230.00250.0028
    下载: 导出CSV
  • [1]

    Hege E K, Beckers J M, Strittmatter P A, McCarthy D W 1985 Appl. Opt. 24 2565Google Scholar

    [2]

    罗群, 黄林海, 顾乃庭, 李斐, 饶长辉 2012 物理学报 61 069501Google Scholar

    Luo Q, Huang L H, Gu N T, Li F, Rao C H 2012 Acta Phys. Sin. 61 069501Google Scholar

    [3]

    颜召军, 陈欣扬, 郑立新, 丁媛媛, 朱能鸿 2016 物理学报 65 199501Google Scholar

    Yan Z J, Chen X Y, Zheng L X, Ding Y Y, Zhu N H 2016 Acta Phys. Sin. 65 199501Google Scholar

    [4]

    常军, 张正慧, 王蕊瑞 2011 物理学报 60 034218Google Scholar

    Chang J, Zhang Zh H, Wang R R 2011 Acta Phys. Sin. 60 034218Google Scholar

    [5]

    Platt B C, Shack R 2001 J. Refract. Surg. 17 S573Google Scholar

    [6]

    Harvey J E, Rockwell R A 1988 Opt. Eng. 27 762Google Scholar

    [7]

    Baron F, Cassaing F, Blanca A, Laubier D 2003 Opt. Eng. 4852 663Google Scholar

    [8]

    Toni F 2010 Phys. Today. 63 26Google Scholar

    [9]

    Chanan G, Troy M, Dekens F, Michaels S, Nelson J, Mast T, Kirkman D 1998 Appl. Opt. 37 140Google Scholar

    [10]

    Shi F, Chanan G, Ohara C, Troy M, David C 2004 Appl. Opt. 43 4474Google Scholar

    [11]

    Li D, Xu S Y, Wang D, Yan D J 2019 Opt. Lett. 44 1170Google Scholar

    [12]

    Pizarro C, Arasa J, Laguarta F, Tomàs N 2002 Appl. Opt. 41 4562Google Scholar

    [13]

    Roddier F 1990 Appl. Opt. 29 1402Google Scholar

    [14]

    Dohlen K, Langlois M, Lanzoni P, Mazzanti S, Vigan A, Montoya L, Hernandez E, Reyes M, Surdej I, Yaitskova N 2006 Proc. SPIE 6267 626734Google Scholar

    [15]

    Esposito S, Pinna E, Puglisi A, Tozzi A, Stefanini P 2005 Opt. Lett. 30 2572Google Scholar

    [16]

    Pinna E, Esposito S, Puglisi A, Pieralli F, Myers R M, Busoni L, Tozzi A, Stefanini P 2006 Proc. SPIE 6267 62672YGoogle Scholar

    [17]

    Wang S S, Zhu Q D, Zhao W R, Li L, Cao G R 2009 Chin. Opt. Lett. 07 1007

    [18]

    Jiang A M, Wang S, Dong Z C, Xue J W, Wang J Y, Dai Y F 2018 Appl. Opt. 57 2736Google Scholar

    [19]

    Salinas-Luna J, Luna E, Salas L, Cruz-González I, Cornejo-Rodríguez A 2004 Opt. Express 12 3719Google Scholar

    [20]

    Vasishta G, Bootha A J, Colavitaa M M, Johnsona R L, Ligona E R, Moore J D, Palmera D L 2003 Proc. SPIE 4838 824Google Scholar

    [21]

    Booth J A, Adams M T, Ames G H, Fowler J R, Rakoczy J M 2000 Proc. SPIE 4003 176Google Scholar

    [22]

    易红伟, 李英才, 樊超, 王矫 2008 光子学报 37 1373

    Yi H W, Li Y C, Fan C, Wang J 2008 Acta Photon. Sin. 37 1373

    [23]

    Dailos G R, Lara D G, Juan T S, Jose R M 2018 Opt. Lett. 43 4264Google Scholar

    [24]

    Hui M, Li W Q, Liu M, Dong L Q, Kong L Q, Zhao Y J 2020 Appl. Opt. 59 771Google Scholar

    [25]

    Ma X F, Xie Z L, Ma H T, Xu Y J, Ren G, Liu Y 2019 Opt. Express 27 16058Google Scholar

    [26]

    Clampin M 2008 Proc. SPIE 41 254Google Scholar

    [27]

    Cheetham A C, Tuthill P G, Sivaramakrishnan A, Lloyd J P 2012 Opt. Express 20 29457Google Scholar

    [28]

    Cheetham A, Cvetojevic N, Norris B, Sivaramakrishnan A, Tuthill P 2014 Opt. Express 22 12924Google Scholar

  • [1] 朱琦, 许多, 张元军, 李玉娟, 王文, 张海燕. 基于卷积神经网络的白蚀缺陷超声探测. 物理学报, 2022, 71(24): 244301. doi: 10.7498/aps.71.20221504
    [2] 张航, 胡月姣, 陈嘉文, 修龙汪. 程能映射下配光平移群的深度神经网络实现. 物理学报, 2022, 71(13): 134201. doi: 10.7498/aps.71.20220178
    [3] 战庆亮, 白春锦, 葛耀君. 基于时程深度学习的复杂流场流动特性表征方法. 物理学报, 2022, 71(22): 224701. doi: 10.7498/aps.71.20221314
    [4] 隋怡晖, 郭星奕, 郁钧瑾, Alexander A. Solovev, 他得安, 许凯亮. 生成对抗网络加速超分辨率超声定位显微成像方法研究. 物理学报, 2022, 71(22): 224301. doi: 10.7498/aps.71.20220954
    [5] 魏连锁, 李华, 吴迪, 郭媛. 基于BP神经网络模型时钟同步误差补偿算法. 物理学报, 2021, 70(11): 114203. doi: 10.7498/aps.70.20201641
    [6] 黄伟建, 李永涛, 黄远. 基于混合神经网络和注意力机制的混沌时间序列预测. 物理学报, 2021, 70(1): 010501. doi: 10.7498/aps.70.20200899
    [7] 周静, 张晓芳, 赵延庚. 一种基于图像融合和卷积神经网络的相位恢复方法. 物理学报, 2021, 70(5): 054201. doi: 10.7498/aps.70.20201362
    [8] 徐启伟, 王佩佩, 曾镇佳, 黄泽斌, 周新星, 刘俊敏, 李瑛, 陈书青, 范滇元. 基于深度卷积神经网络的大气湍流相位提取. 物理学报, 2020, 69(1): 014209. doi: 10.7498/aps.69.20190982
    [9] 王晨阳, 段倩倩, 周凯, 姚静, 苏敏, 傅意超, 纪俊羊, 洪鑫, 刘雪芹, 汪志勇. 基于遗传算法优化卷积长短记忆混合神经网络模型的光伏发电功率预测. 物理学报, 2020, 69(10): 100701. doi: 10.7498/aps.69.20191935
    [10] 颜召军, 陈欣扬, 郑立新, 丁媛媛, 朱能鸿. 基于色散干涉图像的拼接望远镜共相零位标定方法研究. 物理学报, 2016, 65(19): 199501. doi: 10.7498/aps.65.199501
    [11] 周亮, 刘朝晖, 折文集. 可调谐相位板空域频域联合分析. 物理学报, 2015, 64(22): 224207. doi: 10.7498/aps.64.224207
    [12] 李鑫楠, 黄贺艳, 贾小宁, 马驷良. 基于指导滤波的图像盲复原算法. 物理学报, 2015, 64(13): 134202. doi: 10.7498/aps.64.134202
    [13] 李娜, 贾迪, 赵慧洁, 苏云, 李妥妥. 基于改进维纳逆滤波的衍射成像光谱仪数据误差分析与重构. 物理学报, 2014, 63(17): 177801. doi: 10.7498/aps.63.177801
    [14] 王华英, 于梦杰, 江亚男, 宋修法, 朱巧芬, 刘飞飞. 利用小尺寸电荷耦合器件实现数字全息高分辨成像. 物理学报, 2013, 62(24): 244203. doi: 10.7498/aps.62.244203
    [15] 石明珠, 许廷发, 梁炯, 李相民. 单幅模糊图像点扩散函数估计的梯度倒谱分析方法研究. 物理学报, 2013, 62(17): 174204. doi: 10.7498/aps.62.174204
    [16] 罗群, 黄林海, 顾乃庭, 李斐, 饶长辉. 相位差波前检测方法应用于平移误差检测的实验研究. 物理学报, 2012, 61(6): 069501. doi: 10.7498/aps.61.069501
    [17] 赵贵敏, 陆明珠, 万明习, 方莉. 高分辨率扇形阵列超声激发振动声成像研究. 物理学报, 2009, 58(9): 6596-6603. doi: 10.7498/aps.58.6596
    [18] 刘光杰, 单 梁, 戴跃伟, 孙金生, 王执铨. 基于混沌神经网络的单向Hash函数. 物理学报, 2006, 55(11): 5688-5693. doi: 10.7498/aps.55.5688
    [19] 孔祥龙, 李玉同, 远晓辉, 于全芝, 郑志远, 梁文锡, 王兆华, 魏志义, 张 杰. Lucy-Richardson算法用于针孔图像的恢复. 物理学报, 2006, 55(5): 2364-2370. doi: 10.7498/aps.55.2364
    [20] 陈法新, 郑 坚, 杨建伦. 中子厚针孔成像数值模拟研究. 物理学报, 2006, 55(11): 5947-5952. doi: 10.7498/aps.55.5947
计量
  • 文章访问数:  3815
  • PDF下载量:  55
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-03-10
  • 修回日期:  2022-04-27
  • 上网日期:  2022-08-10
  • 刊出日期:  2022-08-20

/

返回文章
返回