-
非对称共光路相干色散光谱仪(coherent-dispersion spectrometer, CODES)是一种基于视向速度法的系外行星探测仪器,通过测量恒星吸收线干涉光谱的多普勒相移探测视向速度的变化。然而恒星吸收谱线中背景白光对CODES相位解析产生干扰,从而严重影响视向速度探测精度。针对背景白光干扰问题,本文利用CODES原理及其探测数据特点,基于U-Net架构提出了背景白光预测网络模型(Background White light Prediction Network, BWP-Net)。该模型先通过结合多通道卷积和深度可分离卷积,从恒星吸收线干涉光谱逐步提取不同级别特征,再通过多层注意力反卷积,融合深层特征和浅层特征基础上逐步重建图像细节,最终预测输出背景白光干涉光谱。实验结果表明,在不同吸收线、不同固定光程差、不同视向速度条件下,利用BWP-Net模型输出消除背景白光干扰后,视向速度探测误差均低于1m/s,误差范围主要集中在0~0.4m/s。该模型不仅能够准确预测背景白光,且具有较强的稳定性和鲁棒性,为CODES高精度稳定探测视向速度提供有力保障。Coherent-dispersion spectrometer (CODES) is an exoplanet detection instrument based on the radial velocity (RV) method. It detects changes in RV by measuring the Doppler phase shift of the interference spectrum of stellar absorption line. However, the background white light in the stellar absorption spectrum makes disturbance to the phase analysis of CODES, which leads to phase error and seriously affects the accuracy of RV inversion. The larger the cosine amplitude of the background white light, the greater the error. To remove background white light effectively for correct Doppler phase shift, a Background White light Prediction Network (BWP-Net) is proposed based on the U-Net architecture in this paper, by utilizing the principle and data characteristics of CODES. To accelerate the convergence of the BWP-Net model, the interference spectrum of absorption line from CODES and the ideal interference spectrum of background white light are used as inputs and labels for the model after image normalization, while the model output becomes the predicted interference spectrum of background white light after inverse normalization. BWP-Net consists of symmetric 6-layer encoding path and decoding path. First, in the encoding path, different levels of features are extracted step by step from the interference spectrum of stellar absorption line through combination of multi-channel convolution and depthwise separable convolution, extracting features effectively while reducing computational costs reasonably. In each convolution layer, spatial downsampling is performed through convolution with a stride of 2 and the number of feature channels are increased until the fourth layer, thus various features, from simple to abstract, local to global, are extracted for preparation of image reconstruction in the decoding path. Second, in the decoding path, the image details are gradually reconstructed from the features extracted through several layers of Attention Transposed-convolution. In each layer of Attention Transposed-convolution, spatial upsampling is performed based on the fusion of shallow features and deep features through matrix addition and the number of feature channels are decreased, while features are given different levels of attention by a learnable weight matrix, so as to suppress absorption line information gradually during image reconstruction. At the last layer of the decoding path, sigmoid activation function is used to control the model output within the 0-1 interval, making it easier to denormalize. Finally, training is performed with region weighted loss function, which combines Mean-Square Error and Multi-Scale Structural Similarity, to consider both of the pixel level differences and structural similarity between the model output and the label, while enhances the suppression of absorption lines in the central region of the interference spectrum through region weighting. And the output of BWP-Net is the prehdiction of the interference spectrum of background white light, which is subtracted from the interference spectrum of stellar absorption lines for phase analysis. The experimental results show that under different absorption lines, different fixed optical path differences, and different RV, after removing background white light with the output of BWP-Net, the RV inversion error is less than 1m/s and mainly concentrated in 0-0.4m/s, while mean error is 0.2353m/s and root mean square error is 0.3769m/s. And the distribution of RV inversion error is relatively uniform under different parameter conditions, the median error is less than 0.25m/s at different absorption line wavelengths, and less than 0.2m/s at different fixed optical path differences. This indicates that BWP-Net not only predicts background white light accurately, but also has good stability and robustness, providing strong support for high-precision and stable RV inversion for CODES.
-
Keywords:
- Radial Velocity /
- Doppler phase shift /
- Background white light prediction /
- Convolutional neural network
-
[1] Bailey I J, Mateo M, White J R, Shectman A S, Crane D J 2018MNRAS 475 1609
[2] Vanzi L, Zapata A, Flores M, Brahm R, Pinto T M, Rukdee S, Jones M, Ropert S, Shen T, Ramirez S, Suc V, Jordán A, Espinoza N 2018MNRAS 477 5041
[3] Mamajek E E, Burgasser J A 2025 AJ 169 77
[4] Laliotis K, Burt A J, Mamajek E E, Li Z, Perdelwitz V, Zhao J, Butler P R, Holden B, Rosenthal L, Fulton J B, Feng F, Kane R S, Bailey J, Carter B, Crane D J, Furlan E , Gnilka L C, Howell B S, Laughlin G, Shectman A S, Teske K J, Tinney G C, Vogt S S, Wang X S, Wittenmyer A R 2023AJ 165 176
[5] The Extrasolar Planets Encyclopedia [2025-1-17]
[6] Wang X, Chang L, Wang L, Ji H, Xian H, Tang Z, Xin Y, Wang C, He S, Zhang J, Lun B, Wei K, Li X, Jiang X, Wang H, Li H, Mao J 2020Res. Astron. Astrophys. 20 032
[7] Xiao G, Teng H, Zhou J, Sato B, Liu Y, Bi S, Takarada T, Kuzuhara M, Hon M, Wang L, Omiya M, Harakawa H, Zhao F, Zhao G, Kambe E, Izumiura H, Ando H, Noguchi K, Wang W, Zhai M, Song N, Yang C, Li T, Brandt D T, Yoshida M, Yoichi Itoh, Kokubo E 2024AJ 167 59
[8] Luo X, Gu S, Xiang Y, Cameron A C, Kim K, Han I, Lee B 2022AJ 163 287
[9] Wang C, Bai J, Fan Y, Mao J, Chang L, Xin Y, Zhang J, Lun B, Wang J, Zhang X, Ying M, Lu K, Wang X, Ji K, Xiong D, Yu X, Ding X, Ye K, Xing L, Yi W, Xu L, Zheng X, Feng Y, He S, Wang X, Liu Z, Chen D, Xu J, Qin S, Zhang R, Tan H, Li Z, Lou K, Li J, Liu W 2019Res. Astron. Astrophys. 19 149
[10] Grieves N, Ge J, Thomas N, Ma B, Sithajan S, Ghezzi L, Kimock B, Willis K, Lee D N, Brian Lee, Fleming W S, Agol E, Troup N, Paegert M, Schneider P D, Stassun K, Varosi F, Zhao B, Jian L, Li R, Mello P F G, Bizyaev D, Pan K, Dutra-Ferreira L, Lorenzo-Oliveira D, Santiago X B, Costa N L, Maia G A M, Ogando C L R, Peloso F E 2017 MNRAS 467 4264
[11] Grieves N, Ge J, Thomas N, Willis K, Ma B, Lorenzo-Oliveira D, Queiroz A B A, Ghezzi L, Chiappini C, Anders F, Dutra-Ferreira L, Mello P F G, Santiago X B, Costa N L, Ogando C L R, Peloso F E, Tan C J, Schneider P D, Pepper J, Stassun G K, Zhao B, Bizyaev D, Pan K 2018 MNRAS 481 3244
[12] Wei R, Chen S, Hu B, Yan Q, Wu Y, Wang P 2020PASP 132 015003
[13] Chen S, Wei R, Xie Z, Wu Y, Di L, Wang F, Zhai Y 2021Appl. Opt. 60 4535
[14] Guan S, Liu B, Chen S, Wu Y, Wang F, Liu X, Wei R 2024Sci Rep 14 17445
[15] Wu Y, Chen S, Wang P, Zhou S, Feng Y, Zhang W, Wei R 2021MNRAS 503 3032
[16] Guan S, Liu B, Chen S, Wu Y, Wang F, Wang S, Liu X, Wei R 2024Opt. Commun. 561 130443
[17] Zhou J, Zhang X, Zhao Y 2021Acta Phys. Sin. 70 054201(in Chinese)
[18] Zhu Q, Xu D, Zhang Y, Li Y, Wang W, Zhang H 2022Acta Phys. Sin. 71 244301(in Chinese)
[19] Long J, Shelhamer E, Darrell T 2015 arXiv:1411.4038v2[cs.CV]
[20] Roy K S, Krishna G, Dubey R S, Chaudhuri B B 2020IEEE Geosci. Remote S. 17 277
[21] Ronneberger O, Fischer P, Brox T 2015Medical Image Computing and Computer -Assisted Intervention Munich, Germany, October 5-9, 2015 p234
[22] Nehaa F, Bhatia D, Shuklab K D, Dalvia M S, Mantzouc N, Shubbar S 2024 arXiv: 2412.02242v1[eess.IV]
[23] Siddique N, Sidike P, Elkin C, Devabhaktuni V 2020 arXiv:2011.01118[eess.IV]
[24] Isola P, Zhu J, Zhou T, Efros A A 2018 arXiv:1611.07004v3[cs.CV]
[25] Basu A, Mondal R, Bhowmik S, Sarkara R 2020J. Electron. Imaging 29 063019
[26] Hu Y, Tang Z, Hu J, Lu X, Zhang W, Xie Z, Zuo H, Li L, Huang Y 2023Opt. Commun.540 129488
计量
- 文章访问数: 63
- PDF下载量: 2
- 被引次数: 0