Abstract


In planetary defense, long period comets have remained a class by themselves. They are recognized as the potentially most devastating threat (i.e. “extinction level events”). But any new comet discovered on an impact trajectory would likely only be discovered as it passed Jupiter, just a few years before impact. However, with machine learning, meteor showers may offer a clue. The proposal: replace the data analyst in the ‘CAMS’ meteor shower survey program by deep learning algorithms and thus enable a global expansion and temporal coverage of a camera network that can detect the dust trails of those potentially hazardous long period comets that came close to Earth’s orbit in the past ten millennia.

The goal is to add years of extra warning time by providing comet searchers directions on where to look for comets when they are still far out. This task is particularly suited to a machine learning approach because of the large scale of data, the need for integration of surveys from around the globe without human intervention, and the need to operate for a long period of time.

The deep learning algorithms would be used to recognize meteors amongst false positives (e.g., satellites), and can triangulate the meteor trajectory in Earth’s atmosphere, its entry speed, and the pre-impact orbit in space through combining different camera perspectives to the same meteor.

Researchers

Andres Plata Stapper*
Antonio Ordonez*
Jack Collison*
Marcello di Cicco*
Susana Zoghbi*

* = Equal contribution

Mentors

Highlights

  • Celestial Globe Web Portal
    A new web portal for CAMS data visualization has come online. It is now possible to see last night's meteor detections by the LOCAMS and Florida networks and the most recently reduced data from the BeNeLux network by Martin Breukers and Carl Johannink. Just go to http://cams.seti.org/FDL/ and choose the date from the "Pick a date" window. Data from the other CAMS networks will gradually come online in the coming months. This new web tool is part of an effort by Frontier Development Lab 2017, a NASA research accelerator program at the SETI Institute, led by James Parr and Bill Diamond and supported by NVidia and IBM amongst others, that set out to use artificial intelligence techniques to automate the CAMS data reduction pipeline. The new web tool designed by Jenniskens, Ganju and Leo Silverberg displays each shower radiant in sun-centered ecliptic coordinates, with a color assigned proportional to the entry speed. By hovering the cursor over a colored meteor radiant, one can see the IAU shower number. Clicking brings up a new window that shows the 2010-2016 CAMS data for that shower displayed in the planetarium program by Ian Webster. That makes the new web tool a portal to the minor showers in Webster's visualization program.
  • Automation
    Python scripts designed by Jack Collison, Peter Jenniskens, and Siddha Ganju collect the submitted files, run the CAMS software, and compare the calculated orbit with a meteor shower template file based on showers assigned by Jenniskens. Susana Zoghbi, Antonio Ordonez, Marcelo de Cicco, and Pete Gural developed machine learning and deep-learning tools to discriminate meteors from other detections. Gural will use this to improve future versions of the Confirmation and Coincidence programs. Further improvements are expected now Jim Albers, Dave Samuels, Steve Rau and Peter Jenniskens are upgrading the hardware and software environment at the existing CAMS networks. The CAMS data reduction automation will make it possible to expand and create new low-light video camera networks to facilitate a continuous global monitoring of night-time meteor shower activity.
  • Discovery
    Andres Plata Stapper used deep-learning tools to discover additional structure in the 2010-2016 shower data pre-assigned by Jenniskens, which in the near future is expected to improve and enrich the meteor shower assignment template. Also, David Holman is working to created a new stream finder tool to help improve the shower isolation from the sporadic background.
  • FDL 2017 Handbook
  • Long Period Comets Mission
  • Long Period Comets Team Photograph

Presented at

@inproceedings{CiccoIMC17,
author = {Marcello De Cicco and S. Zoghbi and A. P. Stapper and A. J. Ordonez and J. Collison and P. S. Gural and S. Ganju and J.-L. Galache and P. Jenniskens},
title = {Artificial Intelligence Techniques applied to Automating Meteor Validation and Trajectory Quality Control to Direct the Search for Long Period Comets},
booktitle = {IMC},
year = {2017}
}
[International Meteor Conference 2017] [Marcello's presentation] [Media]


@inproceedings{ZoghbiGTC17,
author = {S. Zoghbi and M. Cicco and A. P. Stapper and A. J. Ordonez and J. Collison and P. S. Gural and S. Ganju and J.-L. Galache and P. Jenniskens},
title = {In Search of Long-Period Comets: Deep Learning tools for Automatic Meteor Classification and Shower Characterization},
booktitle = {GTC},
year = {2017}
}
[GTC Munich 2017] [Susana's presentation]