arXiv subjects

Martin Cupak

Publications and source records attributed to Martin Cupak.

5 recordsLinked to original sources

arXiv2022

Successful Recovery of an Observed Meteorite Fall Using Drones and Machine Learning

We report the first-time recovery of a fresh meteorite fall using a drone and a machine learning algorithm. A fireball on the 1st April 2021 was observed over Western Australia by the Desert Fireball Network, for which a fall area was calculated for the predicted surviving mass. A search team arrived on site and surveyed 5.1 km2 area over a 4-day period. A convolutional neural network, trained on previously-recovered meteorites with fusion crusts, processed the images on our field computer after each flight. meteorite candidates identified by the algorithm were sorted by team members using two user interfaces to eliminate false positives. Surviving candidates were revisited with a smaller drone, and imaged in higher resolution, before being eliminated or finally being visited in-person. The 70 g meteorite was recovered within 50 m of the calculated fall line using, demonstrating the effectiveness of this methodology which will facilitate the efficient collection of many more observed meteorite falls.

astro-ph.EP↗

arXiv2020

Machine Learning for Semi-Automated Meteorite Recovery

We present a novel methodology for recovering meteorite falls observed and constrained by fireball networks, using drones and machine learning algorithms. This approach uses images of the local terrain for a given fall site to train an artificial neural network, designed to detect meteorite candidates. We have field tested our methodology to show a meteorite detection rate between 75-97%, while also providing an efficient mechanism to eliminate false-positives. Our tests at a number of locations within Western Australia also showcase the ability for this training scheme to generalize a model to learn localized terrain features. Our model-training approach was also able to correctly identify 3 meteorites in their native fall sites, that were found using traditional searching techniques. Our methodology will be used to recover meteorite falls in a wide range of locations within globe-spanning fireball networks.

astro-ph.EP↗

arXiv2019

Recreating the OSIRIS-REx Slingshot Manoeuvre from a Network of Ground-Based Sensors

Optical tracking systems typically trade-off between astrometric precision and field-of-view. In this work, we showcase a networked approach to optical tracking using very wide field-of-view imagers that have relatively low astrometric precision on the scheduled OSIRIS-REx slingshot manoeuvre around Earth on September 22nd, 2017. As part of a trajectory designed to get OSIRIS-REx to NEO 101955 Bennu, this flyby event was viewed from 13 remote sensors spread across Australia and New Zealand to promote triangulatable observations. Each observatory in this portable network was constructed to be as lightweight and portable as possible, with hardware based off the successful design of the Desert Fireball Network. Over a 4 hour collection window, we gathered 15,439 images of the night sky in the predicted direction of the OSIRIS-REx spacecraft. Using a specially developed streak detection and orbit determination data pipeline, we detected 2,090 line-of-sight observations. Our fitted orbit was determined to be within about 10~km of orbital telemetry along the observed 109,262~km length of OSIRIS-REx trajectory, and thus demonstrating the impressive capability of a networked approach to SSA.

astro-ph.IM↗

arXiv2019

Fireball streak detection with minimal CPU processing requirements for the Desert Fireball Network data processing pipeline

The detection of fireballs streaks in astronomical imagery can be carried out by a variety of methods. The Desert Fireball Network--DFN--uses a network of cameras to track and triangulate incoming fireballs to recover meteorites with orbits. Fireball detection is done on-camera, but due to the design constraints imposed by remote deployment, the cameras are limited in processing power and time. We describe the processing software used for fireball detection under these constrained circumstances. A cascading approach was implemented, whereby computationally simple filters are used to discard uninteresting portions of the images, allowing for more computationally expensive analysis of the remainder. This allows a full night's worth of data; over 1000 36 megapixel images to be processed each day using a low power single board computer. The algorithms chosen give a single camera successful detection large fireball rate of better than 96 percent, when compared to manual inspection, although significant numbers of false positives are generated. The overall network detection rate for triangulated large fireballs is estimated to be better than 99.8 percent, by ensuring that there are multiple double stations chances to detect one fireball.

astro-ph.IM↗

arXiv2018

3D Meteoroid Trajectories

Meteoroid modelling of fireball data typically uses a one dimensional model along a straight line triangulated trajectory. The assumption of a straight line trajectory has been considered an acceptable simplification for fireballs, but it has not been rigorously tested. The unique capability of the Desert Fireball Network (DFN) to triangulate discrete observation times gives the opportunity to investigate the deviation of a meteoroid's position to different model fits. Here we assess the viability of a straight line assumption for fireball data in two meteorite-dropping test cases observed by the Desert Fireball Network (DFN) in Australia -- one over 21 seconds (\textit{DN151212\_03}), one under 5 seconds (\textit{DN160410\_03}). We show that a straight line is not valid for these two meteorite dropping events and propose a three dimensional particle filter to model meteoroid positions without any straight line constraints. The single body equations in three dimensions, along with the luminosity equation, are applied to the particle filter methodology described by \citet{Sansom2017}. Modelling fireball camera network data in three dimensions has not previously been attempted. This allows the raw astrometric, line-of-sight observations to be incorporated directly. In analysing these two DFN events, the triangulated positions based on a straight line assumption result in the modelled meteoroid positions diverging up to $3.09\, km$ from the calculated observed point (for \textit{DN151212\_03}). Even for the more typical fireball event, \textit{DN160410\_03}, we see a divergence of up to $360$\,m. As DFN observations are typically precise to $<100$\,m, it is apparent that the assumption of a straight line is an oversimplification that will affect orbit calculations and meteorite search regions for a significant fraction of events.

astro-ph.EP↗