Abstract
Further advances in exoplanet detection and characterisation require sampling a diverse population of extrasolar planets. One technique to detect these distant worlds is through the direct detection of their thermal emission. The so-called direct imaging technique, is suitable for observing young planets far from their star. These are very low signal-to-noise-ratio (SNR) measurements and limited ground truth hinders the use of supervised learning approaches. In this paper, we combine deep generative and discriminative models to bypass the issues arising when directly training on real data. We use a Generative Adversarial Network to obtain a suitable dataset for training Convolutional Neural Network classifiers to detect and locate planets across a wide range of SNRs. Tested on artificial data, our detectors exhibit good predictive performance and robustness across SNRs. To demonstrate the limits of the detectors, we provide maps of the precision and recall of the model per pixel of the input image. On real data, the models can re-confirm bright source detections.
Keywords
- Exoplanet detection
- Direct imaging
- Computer vision
- Generative Adversarial Networks
- Convolutional Neural Networks
This is a preview of subscription content, access via your institution.
Buying options
Notes
- 1.
Paris Observatory Exoplanet Catalogue: http://exoplanet.eu.
- 2.
The term flux refers to the rate of incoming photons.
- 3.
There are no confirmed planet detections on NICMOS filter F110W yet. These bright sources are almost certainly background stars. However, detecting these showcases the potential for any bright source –including planets– to be detected.
- 4.
The original dataset is publicly available at the HST LAPLACE STScI archive https://archive.stsci.edu/prepds/laplace/.
- 5.
The Field of View (FOV) of the camera, is 19.2\({^\prime }{^\prime }\) \(\times \,\)19.2\({{^\prime }{^\prime }}\) corresponding to images of size 256 \(\times \) 256 pixels and the coronagraph is a circular disk with a radius of 4 pixels.
- 6.
Bias calibration removes unwanted saturated pixels that arise during long exposures. Dark calibration corrects for thermal emissions coming from the detector. Flat calibration corrects for differences in sensitivity across the CCD detector.
- 7.
Large bright spots found outside the speckle pattern.
- 8.
Data and code are available at https://github.com/ucl-exoplanets/DI-Project.
- 9.
The term refers to artifacts caused by the uneven overlap of the deconvolutions of a CNN when the kernel (filter) size is not divisible by the number of strides.
- 10.
- 11.
We opted not to use a GAN for augmenting the positive examples (i) to fully control the SNR of the injected planets, for evaluation purposes and (ii) because the randomly positioned faint planet signal in positive examples would be easily masked by the most prevalent features of the images, i.e. those comprising the speckle pattern.
- 12.
The PSF is the response of the telescope optics to incoming light, i.e. it defines the light distribution of a point-source, e.g. a planet, on the detector plane.
- 13.
The window’s top-left corner is \((x-1, y-1)\) and bottom-right is \((x+2, y+2)\).
- 14.
References
Amara, A., Quanz, S.P.: PYNPOINT: an image processing package for finding exoplanets. MNRAS 427, 948–955 (2012)
Bowler, B.P.: Imaging extrasolar giant planets. Publ. Astron. Soc. Pac. 128, 102001 (2016)
Cantalloube, F., et al.: Direct exoplanet detection and characterization using the andromeda method: performance on VLT/NaCo data. Astron. Astrophys. 582, A89 (2015)
Cassan, A., et al.: One or more bound planets per Milky Way star from microlensing observations. Nature 481, 167–169 (2012)
Choquet, É., et al.: HD 104860 and HD 192758: two debris disks newly imaged in scattered light with the hubble space telescope. Astrophys. J. 854(1), 53 (2018)
Fergus, R., Hogg, D.W., Oppenheimer, R., Brenner, D., Pueyo, L.: S4: a spatial-spectral model for speckle suppression. Astrophys. J. 794(2), 161 (2014)
Gomez Gonzalez, C.A., Absil, O., Van Droogenbroeck, M.: Supervised detection of exoplanets in high-contrast imaging sequences. Astron. Astrophys. 613, A71 (2018)
Gomez Gonzalez, C.A., et al.: Low-rank plus sparse decomposition for exoplanet detection in direct-imaging ADI sequences. The LLSG algorithm. Astron. Astrophys. 589, A54 (2016)
Goodfellow, I.J., et al.: Generative adversarial networks. ArXiv (2014)
Hagan, J.B., Choquet, É., Soummer, R., Vigan, A.: ALICE data release: a revaluation of HST-NICMOS coronagraphic images. Astron. J. 155, 179 (2018)
Kalas, P., et al.: Optical images of an exosolar planet 25 light-years from earth. Science 322, 1345 (2008)
Krist, J.E., Hook, R.N., Stoehr, F.: 20 years of Hubble Space Telescope optical modeling using Tiny Tim. In: Optical Modeling and Performance Predictions V. Proceedings of the Society of Photo-Optical Instrumentation Engineers, vol. 8127, p. 81270J (2011)
Lafrenière, D., Marois, C., Doyon, R., Nadeau, D., Artigau, É.: A new algorithm for point-spread function subtraction in high-contrast imaging: a demonstration with angular differential imaging. Astrophys. J. 660, 770–780 (2007)
Lagrange, A.M., et al.: A giant planet imaged in the disk of the young star \(\beta \) Pictoris. Science 329, 57 (2010)
Mugnier, L.M., et al.: Optimal method for exoplanet detection by angular differential imaging. J. Opt. Soc. Am. A 26, 1326 (2009)
Racine, R., Walker, G.A., Nadeau, D., Doyon, R., Marois, C.: Speckle noise and the detection of faint companions. Publ. Astron. Soc. Pac. 111(759), 587 (1999)
Radford, A., Metz, L., Chintala, S.: Unsupervised representation learning with deep convolutional generative adversarial networks. ArXiv (2015)
Soummer, R., Pueyo, L., Larkin, J.: Detection and characterization of exoplanets and disks using projections on Karhunen-Loève eigenimages. Astrophys. J. Lett. 755, L28 (2012)
Yeh, R.A., et al.: Semantic image inpainting with deep generative models. In: CVPR, pp. 5485–5493 (2017)
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., Torralba, A.: Learning deep features for discriminative localization. In: CVPR, pp. 2921–2929 (2016)
Acknowledgments
This project has received funding from the European Research Council (ERC) under the EU Horizon 2020 research & innovation programme (grant No 758892, ExoAI) and under the EU Seventh Framework Programme (FP7/2007-2013)/ ERC grant No 617119 (ExoLights). Furthermore, we acknowledge funding by the Science & Technology Funding Council (STFC) grants: ST/K502406/1, ST/P000282/1, ST/P002153/1 and ST/S002634/1. We are grateful for the support of the NVIDIA Corporation through the NVIDIA GPU Grant program.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Yip, K.H. et al. (2020). Pushing the Limits of Exoplanet Discovery via Direct Imaging with Deep Learning. In: Brefeld, U., Fromont, E., Hotho, A., Knobbe, A., Maathuis, M., Robardet, C. (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2019. Lecture Notes in Computer Science(), vol 11908. Springer, Cham. https://doi.org/10.1007/978-3-030-46133-1_20
Download citation
DOI: https://doi.org/10.1007/978-3-030-46133-1_20
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-46132-4
Online ISBN: 978-3-030-46133-1
eBook Packages: Computer ScienceComputer Science (R0)
-
Published in cooperation with
http://www.ecmlpkdd.org/
