Euclid：利用深度学习在模拟图像中识别小行星条带
地球与行星天体物理
2023-11-29 v1 天体物理仪器与方法
机器学习
摘要
在ESA Euclid空间望远镜的图像中将有最多150000颗小行星可见，且Euclid的仪器为这些天体提供多波段可见光至近红外测光以及无缝光谱。大多数小行星在图像中呈现为条带。由于图像和小行星数量庞大，需要自动化探测方法。此前测试了一种基于StreakDet软件的非机器学习方法，但其在短和/或暗弱条带上的结果并非最优。我们着手利用深度学习提升Euclid图像中小行星条带的探测能力。我们使用模拟Euclid图像构建、训练并测试了一个三步机器学习流水线。首先，卷积神经网络(CNN)在全图中探测条带及其坐标，旨在最大化探测的完整度（召回率）。随后，循环神经网络(RNN)将由CNN在若干部分中探测到的长条带片段进行合并。最后，梯度提升树(XGBoost)在不同Euclid曝光之间关联已探测条带，以减少假阳性数量并提升样本的纯度（精确率）。该深度学习流水线超越了基于StreakDet软件的非机器学习流水线的完整度，并达到相似的纯度水平。此外，深度学习流水线可探测比StreakDet暗0.25–0.5星等的小行星。相比StreakDet软件，该深度学习流水线可使探测到的小行星数量增加50%。仍有进一步改进空间，特别是在提升条带坐标精度以及增强流水线最终阶段（涉及跨多次曝光关联探测）的完整度方面。
引用
@article{arxiv.2310.03845,
title = {Euclid: Identification of asteroid streaks in simulated images using deep learning},
author = {M. Pöntinen and M. Granvik and A. A. Nucita and L. Conversi and B. Altieri and B. Carry and C. M. O'Riordan and D. Scott and N. Aghanim and A. Amara and L. Amendola and N. Auricchio and M. Baldi and D. Bonino and E. Branchini and M. Brescia and S. Camera and V. Capobianco and C. Carbone and J. Carretero and M. Castellano and S. Cavuoti and A. Cimatti and R. Cledassou and G. Congedo and Y. Copin and L. Corcione and F. Courbin and M. Cropper and A. Da Silva and H. Degaudenzi and J. Dinis and F. Dubath and X. Dupac and S. Dusini and S. Farrens and S. Ferriol and M. Frailis and E. Franceschi and M. Fumana and S. Galeotta and B. Garilli and W. Gillard and B. Gillis and C. Giocoli and A. Grazian and S. V. H. Haugan and W. Holmes and F. Hormuth and A. Hornstrup and K. Jahnke and M. Kümmel and S. Kermiche and A. Kiessling and T. Kitching and R. Kohley and M. Kunz and H. Kurki-Suonio and S. Ligori and P. B. Lilje and I. Lloro and E. Maiorano and O. Mansutti and O. Marggraf and K. Markovic and F. Marulli and R. Massey and E. Medinaceli and S. Mei and M. Melchior and Y. Mellier and M. Meneghetti and G. Meylan and M. Moresco and L. Moscardini and E. Munari and S. -M. Niemi and T. Nutma and C. Padilla and S. Paltani and F. Pasian and K. Pedersen and V. Pettorino and S. Pires and G. Polenta and M. Poncet and F. Raison and A. Renzi and J. Rhodes and G. Riccio and E. Romelli and M. Roncarelli and E. Rossetti and R. Saglia and D. Sapone and B. Sartoris and P. Schneider and A. Secroun and G. Seidel and S. Serrano and C. Sirignano and G. Sirri and L. Stanco and P. Tallada-Crespí and A. N. Taylor and I. Tereno and R. Toledo-Moreo and F. Torradeflot and I. Tutusaus and L. Valenziano and T. Vassallo and G. Verdoes Kleijn and Y. Wang and J. Weller and G. Zamorani and J. Zoubian and V. Scottez},
journal= {arXiv preprint arXiv:2310.03845},
year = {2023}
}
备注
18 pages, 11 figures