Machine Learning Assisted NEO Discovery and Polarimetric Characterisation with Astronomical Surveys

B.Y. Irureta-Goyena; D. Dirkx; D. Koschny; G.A. Verdoes Kleijn; L.V.E. Koopmans; M. Granvik; M. Micheli; M. P\"ontinen; O.R. Williams; S. Bagnulo

arxiv: 2604.02999 · v1 · submitted 2026-04-03 · 🌌 astro-ph.IM

Machine Learning Assisted NEO Discovery and Polarimetric Characterisation with Astronomical Surveys

G.A. Verdoes Kleijn , T. Grobler , S.J. Chong , O.R. Williams , M. Micheli , D. Koschny , T. Saifollahi , L.V.E. Koopmans

show 7 more authors

D. Dirkx T. Santana-Ros Y.-Z. Ma M. P\"ontinen S. Bagnulo M. Granvik B.Y. Irureta-Goyena

This is my paper

Pith reviewed 2026-05-13 18:21 UTC · model grok-4.3

classification 🌌 astro-ph.IM

keywords near-earth objectsmachine learningastronomical surveyspolarimetryplanetary defenseserendipitous discoverydata analysis platforms

0 comments

The pith

Machine learning can detect and characterize near-Earth objects in surveys designed for galactic targets.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper describes a research programme to build machine learning algorithms and data platforms that identify near-Earth objects in wide-field astronomical survey images. These surveys are built to study fixed galaxies and distant objects, so NEOs appear as unexpected moving sources that must be extracted serendipitously. The programme targets both faster discovery for planetary defense and polarimetric measurements that reveal surface properties of the objects. If the methods work, existing survey archives become a richer source of NEO data without requiring new dedicated observations.

Core claim

We present our research and development programme for algorithms and digital data analysis platforms for machine learning-assisted NEO discovery and polarimetric characterisation in astronomical surveys.

What carries the argument

Machine learning algorithms and digital data analysis platforms that extract moving solar-system objects and their polarimetric signatures from survey data optimized for stationary celestial sources.

If this is right

Existing survey archives can yield additional NEO detections without new telescope time.
Polarimetric characterisation becomes feasible for a larger sample of NEOs, improving knowledge of their composition.
Planetary defense benefits from higher completeness in the catalog of potentially hazardous objects.
The same platforms can be reused to search for other transient solar-system bodies in archival images.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar machine-learning pipelines could be applied to archival data from other large surveys to recover overlooked solar-system objects.
Survey design teams might incorporate moving-object detection as a standard data-product requirement from the outset.
Cross-institutional teams of astronomers and data scientists can accelerate extraction of scientific value from petabyte-scale sky surveys.

Load-bearing premise

Machine learning models trained on survey data can reliably detect and characterize NEOs even though the surveys are optimized for galactic and extragalactic targets rather than moving solar-system objects.

What would settle it

If tests on real survey data show that the models miss most known NEOs or return too many false positives that cannot be filtered efficiently, the claim that these platforms are useful for NEO work would be falsified.

read the original abstract

We are a group of over two dozen astronomers, computer scientists, data scientists and digital Big Data research platform experts at 11 universities and research institutes in South Africa and Europe. We study Near-Earth Objects (NEOs) for Planetary Defence and scientific purposes. We present our research and development programme for algorithms and digital data analysis platforms for machine learning-assisted NEO discovery and polarimetric characterisation in astronomical surveys. Typically, this is serendipitous because these surveys are designed for galactic and extragalactic science.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a program announcement for an ML-assisted NEO discovery effort on serendipitous survey data, with no methods, data, or results presented.

read the letter

The paper's main point is that a large team across South Africa and Europe is starting an R&D program to apply machine learning to existing astronomical surveys for near-Earth object discovery and polarimetry. The surveys target galaxies and extragalactic sources, so the work would rely on serendipitous detections of moving solar-system objects. They correctly flag that the data were not collected with NEOs in mind and that ML will be needed to handle the mismatch in cadence, depth, and selection effects. The team mix of astronomers, data scientists, and platform builders is a reasonable setup for the data-handling side of the problem. That is the extent of what is new here: a stated intention to build algorithms and platforms for this repurposing task. No equations, no training sets, no validation numbers, and no comparison to existing NEO pipelines appear in the text. The central claim is simply that such a program exists and will address planetary defense and solar-system science needs. That claim is accurate as far as it goes, but it carries no empirical weight. The soft spot is exactly that absence of substance. Without even a sketch of the planned ML architecture or a pilot result on real survey frames, there is nothing to assess for soundness or improvement over prior work. Readers already following large-survey data mining or planetary defense coordination might find the team roster and high-level goals useful as a statement of intent. Anyone looking for a technical advance or reproducible finding will come away empty. I would not send this for peer review in its current form. It belongs in a project-overview venue or as a white-paper style note rather than a research article.

Referee Report

1 major / 0 minor

Summary. The manuscript presents a collaborative R&D programme involving over two dozen researchers from 11 institutions in South Africa and Europe. It outlines plans to develop machine learning algorithms and digital data analysis platforms for serendipitous discovery and polarimetric characterisation of Near-Earth Objects (NEOs) in astronomical surveys primarily designed for galactic and extragalactic science.

Significance. If the described programme yields effective ML methods that overcome the mismatch between survey optimisation and solar-system object detection, it could meaningfully advance planetary defence by increasing NEO discovery rates and characterisation using existing large-scale datasets, without the need for dedicated surveys.

major comments (1)

[Abstract] Abstract: the central claim is a description of planned algorithms and platforms, yet no specific algorithms, training datasets, validation metrics, performance numbers, or preliminary results are provided. This absence is load-bearing because it prevents any evaluation of whether the proposed ML approaches can reliably detect and characterise NEOs in surveys optimised for other targets.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thoughtful review and for acknowledging the potential significance of our collaborative R&D programme. We address the major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim is a description of planned algorithms and platforms, yet no specific algorithms, training datasets, validation metrics, performance numbers, or preliminary results are provided. This absence is load-bearing because it prevents any evaluation of whether the proposed ML approaches can reliably detect and characterise NEOs in surveys optimised for other targets.

Authors: The manuscript is intentionally structured as an overview of a multi-institutional research and development programme rather than a report of completed algorithmic work. It describes the collaborative framework, scientific objectives, and high-level planned approaches for serendipitous NEO discovery and polarimetric characterisation using existing galactic/extragalactic surveys. Specific algorithm details, training datasets, metrics, and performance results are not included because they do not yet exist; they will be reported in subsequent technical papers as the programme progresses. We have revised the abstract and introduction to more explicitly state the manuscript's scope as a programme description to avoid any misinterpretation. revision: partial

Circularity Check

0 steps flagged

No significant circularity: descriptive R&D programme with no derivations

full rationale

The manuscript presents a research and development programme for ML-assisted NEO discovery and polarimetric characterisation. It contains no equations, fitted parameters, predictive models, or derivation chains. The central claim is the existence and scope of the programme itself, which is forward-looking and does not reduce to any self-referential input, self-citation load-bearing premise, or fitted prediction. No load-bearing steps exist that could exhibit circularity under the enumerated patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no technical details on parameters, axioms, or new entities; the program description rests on the unstated premise that ML can extract NEO signals from non-optimized survey data.

pith-pipeline@v0.9.0 · 5453 in / 1001 out tokens · 35000 ms · 2026-05-13T18:21:45.754375+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We present our research and development programme for algorithms and digital data analysis platforms for machine learning-assisted NEO discovery and polarimetric characterisation in astronomical surveys. Typically, this is serendipitous because these surveys are designed for galactic and extragalactic science.
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

A machine-learning approach was done by Irureta-Goyena et al (2025). They develop a pipeline in which the Convolution Neural Network TernausNet automatically segments a single image.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

Verdoes Kleijn (1) , T

9th IAA Planetary Defense Conference – PDC 2025 5-9 May 2025, Stellenbosch, Cape Town, South Africa IAA-PDC-25-04-147 Machine Learning Assisted NEO Discovery and Polarimetric Characterisation with Astronomical Surveys G.A. Verdoes Kleijn (1) , T. Grobler (2) , S.J. Chong (1,2) , O.R. Williams (1) , M. Micheli (3) , D. Koschny (4,5) , T. Saifollahi (6) , L...

work page 2025
[2]

Our group also includes members of the Rubin Consortium who are interested in taking the lessons learned from Euclid and OmegaCAM to the LSST survey

and the Euclid Mission surveys (Euclid Collaboration 2024). Our group also includes members of the Rubin Consortium who are interested in taking the lessons learned from Euclid and OmegaCAM to the LSST survey. For the OmegaCAM archive, it is estimated that one in twenty NEOs appear in OmegaCAM data at a signal-to-noise ratio higher than 3 (Saifollahi et a...

work page 2024
[3]

test particles

In other words, currently, the majority of NEOs with a predicted 3 < S/N < 10 remain undetected (even after visual inspection). Especially for Planetary Defence purposes, it is relevant to assess whether the failed NEO detections can indicate errors that are unaccounted for (e.g., in their photometric model). These results make it interesting to develop a...

work page 2025
[4]

Mellier, Abdurro’uf, J

and inducers of climate change (Brugger et al., 2017). 5 NEOs and Astronomical Research Data Platforms We want to develop our pipeline for the detection and classification of NEOs (and asteroids in general) as a generic software component that is straightforward to configure for different surveys and can be interfaced or embedded in various astronomical d...

work page arXiv 2017

[1] [1]

Verdoes Kleijn (1) , T

9th IAA Planetary Defense Conference – PDC 2025 5-9 May 2025, Stellenbosch, Cape Town, South Africa IAA-PDC-25-04-147 Machine Learning Assisted NEO Discovery and Polarimetric Characterisation with Astronomical Surveys G.A. Verdoes Kleijn (1) , T. Grobler (2) , S.J. Chong (1,2) , O.R. Williams (1) , M. Micheli (3) , D. Koschny (4,5) , T. Saifollahi (6) , L...

work page 2025

[2] [2]

Our group also includes members of the Rubin Consortium who are interested in taking the lessons learned from Euclid and OmegaCAM to the LSST survey

and the Euclid Mission surveys (Euclid Collaboration 2024). Our group also includes members of the Rubin Consortium who are interested in taking the lessons learned from Euclid and OmegaCAM to the LSST survey. For the OmegaCAM archive, it is estimated that one in twenty NEOs appear in OmegaCAM data at a signal-to-noise ratio higher than 3 (Saifollahi et a...

work page 2024

[3] [3]

test particles

In other words, currently, the majority of NEOs with a predicted 3 < S/N < 10 remain undetected (even after visual inspection). Especially for Planetary Defence purposes, it is relevant to assess whether the failed NEO detections can indicate errors that are unaccounted for (e.g., in their photometric model). These results make it interesting to develop a...

work page 2025

[4] [4]

Mellier, Abdurro’uf, J

and inducers of climate change (Brugger et al., 2017). 5 NEOs and Astronomical Research Data Platforms We want to develop our pipeline for the detection and classification of NEOs (and asteroids in general) as a generic software component that is straightforward to configure for different surveys and can be interfaced or embedded in various astronomical d...

work page arXiv 2017