Automating Asteroid Detection Criteria to Strengthen Citizen Science for Planetary Defense

On 15 February 2013, a ~19m asteroid airburst over Chelyabinsk released ~500 kt, ~30x Hiroshima [1]. 1,600+ injured, 7,200 buildings damaged, no warning available [18]. NASA estimates ~25,000 NEAs 140m or larger [2, 20] (Figure 1).

Kelly Fast (NASA, AAAS Feb 2026) : "What keeps me up at night is the asteroids we don't know about." [19]

Big telescope surveys detect many asteroids but can still miss. Faint detections filter as noise [13]. As of October 2019, 16.6M unmatched detections sat in the MPC Isolated Tracklet File (Figure 2) [21].

Pan-STARRS produces 4 TB / night [4]. Vera Rubin LSST ~20 TB / night, issuing alerts since February 2026 [5].

IASC has 50,000 citizen scientists in 96 countries [6] (Figure 4). As of April 2026, 19,651 preliminary detections are reported, 193 provisionally confirmed by the MPC [24]. Citizen scientists reach ~0.30 magnitudes dimmer than automation [6]. During my August-September 2024 IASC campaign, I confirmed 2024 RH39 and 2024 RX69 [23].

Most new participants do not return after their first session [8]. Across 14M online citizen science accounts, fewer than half remain active past a year, top 10% complete ~79% of the work [26]. Lack of feedback is researched as a key retention barrier (Figure 5) [9].

Research Question:

Can asteroid detection be automated and applied to citizen-science images to match experienced manual review, speed up asteroid discovery, provide feedback, help citizen scientists submit higher-confidence detections, reduce entry-retention barriers, and strengthen citizen science?

My research automates the manual review of Pan-STARRS images that citizen scientists receive through IASC.

1. The Nine Detection Criteria (Figures 6, 7, 8)

I researched nine detection criteria during my manual asteroid discovery process that helped detect 2024 RH39 and 2024 RX69 using Astrometrica software on IASC Pan-STARRS images. These are persistence (candidate appears in at least 3 of the 4 frames), straight-line motion, constant speed, constant brightness (magnitude variation under 1.0), no match in the Minor Planet Center (MPC) catalog, SNR (Signal to Noise Ratio) above 5 sigma, PSF Gaussian fit (RMS below 0.2), FWHM consistency (0.8 to 1.2 arcseconds), and rejecting false signatures.

Automating these criteria first requires finding candidates in the images. I researched six methods.

2. Six Detection Methods Cast a Wide Net (Figure 10)

Standard Tracklet (5 sigma per-frame linking), Shift-and-Stack (SNR gain), Temporal Consistency (forced-measurement at predicted positions), Deep Detection (combined frames with cosmic rays removed), Blind Temporal Coherence (combinations below the per-frame gate, summed across all four frames), and Dipole Search (frame subtraction).

3. Drawing on Three Areas (Figure 8)

I drew from three areas of my experience : astronomy (coordinate alignment, blink comparison, MPC formatting), astrophotography (SNR stacking, sub-pixel tracking, difference imaging), and physics (Kepler's laws, point-spread functions, FWHM measurement).

4. ARIA Pipeline (Figure 9)

ARIA stands for Automated Recognition and Identification of Asteroids. Pipeline : 4 FITS to 6 detection methods to 9 criteria to MPC catalog match to HTML report. ARIA found both MPC-confirmed asteroids from the same images I processed manually.

5. Detection Range (Figure 9)

ARIA's detection range covers Main-Belt and Near-Earth asteroids in Pan-STARRS data within the magnitude limit (about 21.2 [4,11]). All four candidates (RH39, XV22, PT180, RX69) fall inside (Figure 9).

I tested the hypothesis by running ARIA on 7 IASC image sets from my August and September 2024 IASC campaign. Each set holds 4 Pan-STARRS frames about 16 minutes apart [4, 6].

1. Detection on test fields with known asteroids (Figure 11)

Two of the 7 sets contained known asteroids and served as test fields.

ARIA found all 4 known asteroids in those fields.

Which method found each one:

Standard Tracklet caught (267555) PT180, 2024 RH39, and (734270) XV22.

Blind Temporal Coherence method caught 2024 RX69.

2. 2024 RH39 found by Standard Tracklet (Figure 12)

Found in image set KX06_p10.

Apparent magnitude 19.5.

Combined signal-to-noise 35.8 across 4 frames.

ARIA score 80.9 out of 100.

The Minor Planet Center had confirmed the discovery.

Orbit details: semi-major axis 2.65 AU, eccentricity 0.264, inclination 14.35 degrees, period 4.32 years, Main Belt.

3. 2024 RX69 found by Blind Temporal Coherence (Figure 13)

Found in image set KX73_p13.

Apparent magnitude 21.1.

Each frame's signal-to-noise was about 3 to 4, too dim for single-frame detection.

After stacking all 4 frames, combined signal-to-noise reached 19.9.

ARIA score 64.8 out of 100.

The Minor Planet Center had confirmed the discovery.

Orbit details: semi-major axis 3.18 AU, eccentricity 0.245, inclination 15.59 degrees, period 5.67 years, Main Belt [11].

4. Workflow comparison (Figure 14)

Manual review takes about 45 minutes per image set across 11 steps in Astrometrica: open 4 FITS, compare frames, check path, brightness, signal-to-noise, PSF, FWHM, discard false detections, compare with MPC catalog, measure coordinates, submit MPC report.

ARIA finishes the same set in about 6 minutes across 5 steps: load FITS, run 6 methods plus 9 checks plus MPC catalog match, review flagged detections, validate, submit.

The 7.5x speedup automates each individual check.

Can manually review every flagged detection.

ARIA reaches magnitude 21.2 [4, 11], comparable to ATLAS (19.7) [16] and ZTF (20.5) [17], and below MOPS (22 to 24) [13].

5. Detection depth, limits, conclusions (Figure 15)

Detection depth:

Stacking 4 frames lifts above the 5 sigma per-frame gate.

RX69 clears the gate via Blind Temporal Coherence after stacking.

Five known asteroids in the test fields were too dim or stationary for ARIA to find: (88879) at magnitude 21.9, Q7555 at magnitude 20.5, 111725 at magnitude 21.0 (stationary, no motion in 16 minutes), g0419 at magnitude 20.6, and v6557 at magnitude 20.4.

Limits with next steps:

Starter scope (more IASC rounds),

Detection depth (deeper stacking past the magnitude 21.2 [4, 11] limit),

Fast movers (wider linking range),

Crowded fields (better masking),

Pan-STARRS telescope (add ATLAS, CSS, ZTF for 10 to 20 more sets).

Conclusions:

ARIA automates each check,

Stacking frames catches dim asteroids,

ARIA cuts review time about 7.5x.

1. What this result unlocks for planetary defense:

My system recovered 4 of 4 known asteroids in 7 IASC sets tested (Figure 16).

Magnitude 21.2 depth.

DART proved we can deflect but we need to find them first [35]

2. What feedback could unlock for citizen science:

ARIA gives a 0 to 100 confidence score per candidate

7.5x faster review per session (45 min manual now 6 min)

Coordinates for Astrometrica (Figure 17).

Feedback is the retention research points to [7, 8, 9]. The curve shows what changes with feedback after every try.

3. Where ARIA sits that no tool did:

Pan-STARRS MOPS [13], ATLAS [16], and ZTF [17] automate at survey scale with depth limits 22-24, 19.7, and 20.5 magnitude.

Astrometrica is the professional tool 50,000 citizen scientists use for manual review [6].

ARIA fills the citizen-scientists-automation gap that can run on a laptop (Figure 18).

4. Three takeaways:

Same criteria can be applied to other citizen science programs and telescopes (Figure 19).

Stacking 4 photos reaches magnitude 21.2.

What took 45 minutes or more for each image set manually now takes 6 minutes.

5. From ARIA report to numbered asteroid:

ARIA outputs each candidate with a 0 to 100 confidence score and coordinates, ranked HIGH/MEDIUM/LOW (Figure 20).

The citizen scientist verifies in Astrometrica and submits an MPC-format report to IASC.

IASC forwards to MPC: preliminary, provisional once linked, numbered after multi-year arcs.

Widening the footprint to strengthen validation:

ATLAS, Catalina Sky Survey, and ZTF expand ARIA to more telescopes (Figure 21) [16, 17].

More IASC image sets with confirmed asteroids establish ARIA's detection capabilties [24].

Web platform:

Citizen scientists upload IASC images, get ranked potential asteroid candidates in 6 minutes, explore and learn each concept (PSF/FWHM/SNR etc.,) if required, and potentially contribute to MPC Catalog [23, 24] (Figure 22).

Thanks to:

- My parents

- Dr. Patrick Miller (founder, IASC)

- Clara Brenton Public School (My School)

- International Asteroid Search Campaign (IASC)

- Pan-STARRS team, University of Hawaii (IfA)

- Minor Planet Center (IAU / Smithsonian Astrophysical Observatory)

- Herbert Raab (author of Astrometrica)

- Open-source scientific Python (NumPy, SciPy, Astropy, photutils)

- RASC London Centre (Royal Astronomical Society of Canada)

With thanks also to TVSEF and CWSF organizers.

AI ACKNOWLEDGEMENT

Python code for the ARIA pipeline and SVG visual representations are implemented using Claude Code (Anthropic) [12]. The research and design workflow, along with the 6 detection methods from my fields of experience and the 9 detection criteria from my direct experience of manual asteroid discovery, are employed into the implementation through iterative testing, debugging, fixing issues and tuning, and manual verification in Astrometrica for all the results. All prompts and interactions are logged in my research notebook.

[1] Brown, P. G., Assink, J. D., Astiz, L., Blaauw, R., Boslough,   M. B., Borovicka, J., & Krzeminski, Z. (2013). A 500-   kiloton airburst over Chelyabinsk and an enhanced hazard from

small impactors. Nature, 503(7475), 238-241.   https://doi.org/10.1038/nature12741

[2] Harris, A. W., & D'Abramo, G. (2015). The population of near-   Earth asteroids. Icarus, 257, 302-312.   https://doi.org/10.1016/j.icarus.2015.05.004

[3] Granvik, M., Morbidelli, A., Jedicke, R., Bolin, B., Bottke,   W. F., Beshore, E., Vokrouhlicky, D., Nesvorny, D., &   Michel, P. (2018). Debiased orbit and absolute-magnitude  distributions for near-Earth objects. Icarus, 312, 181-207.   https://doi.org/10.1016/j.icarus.2018.04.018

[4] Chambers, K. C., Magnier, E. A., Metcalfe, N., Flewelling, H.   A., Huber, M. E., Waters, C. Z., . & Wainscoat, R. J.   (2016). The Pan-STARRS1 surveys. arXiv preprint arXiv:1612.05560. https://arxiv.org/abs/1612.05560

[5] Ivezic, Z., Kahn, S. M., Tyson, J. A., Abel, B., Acosta, E.,   Allsman, R., .& Zhan, H. (2019). LSST: From science   drivers to reference design and anticipated data products.   The Astrophysical Journal, 873(2), 111.   https://doi.org/10.3847/1538-4357/ab042c

[6] Miller, P., Knapp, M., Welker, A., Sleeper, J., Phelps, V., &   Stern, N. (2024). The International Asteroid Search   Campaign: Methods, impact, and the citizen-science

contribution to minor-planet discovery. Publications of the  Astronomical Society of the Pacific, 136(2), 024502.   https://doi.org/10.1088/1538-3873/ad11a0

[7] Sauermann, H., & Franzoni, C. (2015). Crowd science user  contribution patterns and their implications. Proceedings of  the National Academy of Sciences, 112(3), 679-684.  https://doi.org/10.1073/pnas.1408907112

[8] Crall, A. W., Jordan, R., Holfelder, K., Newman, G. J., Graham, J., & Waller, D. M. (2017). The impacts of an  invasive species citizen science training program on   participant attitudes, behavior, and science literacy.  Journal of Science Communication, 16(1), A01.   https://doi.org/10.22323/2.16010201

[9] West, S. E., & Pateman, R. M. (2016). Recruiting and retaining  participants in citizen science: What can be learned from the  volunteering literature? Citizen Science: Theory and Practice,  1(2), 15. https://doi.org/10.5334/cstp.8

[10] Howell, S. B. (2006). Handbook of CCD astronomy (2nd ed.).  Cambridge University Press.   https://doi.org/10.1017/CBO9780511807909

[11] Veres, P., Jedicke, R., Denneau, L., Wainscoat, R., Holman,  M. J., & Lin, H.-W. (2012). Improved asteroid astrometry and  photometry with trail fitting. Publications of the

Astronomical Society of the Pacific, 124(921), 1197-1207. https://doi.org/10.1086/668616

[12] Anthropic. (2026). Claude Code [software], v2.1.116 (April 2026   build) [AI coding assistant]. https://code.claude.com/

[13] Denneau, L., Jedicke, R., Grav, T., Granvik, M., Kubica, J.,  Milani, A., Veres, P., ... & Wainscoat, R. (2013). The Pan-  STARRS moving object processing system. Publications of the  Astronomical Society of the Pacific, 125(926), 357-395.

https://doi.org/10.1086/670337

[14] Weryk, R. J., Wainscoat, R. J., Bhatt, M., Schultz, A. S. B.,   Ramanjooloo, Y., Fitzsimmons, A., & Chambers, K. C.   (2020). The Pan-STARRS moving object processing system:

Outbound hyperbolic orbits. The Astronomical Journal, 159(5),  193. https://doi.org/10.3847/1538-3881/ab7ebd

[15] Kubica, J., Denneau, L., Grav, T., Heasley, J., Jedicke, R.,   Masiero, J., Milani, A., Moore, A., Tholen, D., & Wainscoat,   R. J. (2007). Efficient intra- and inter-night linking of  asteroid detections using kd-trees. Icarus, 189(1), 151-168.  https://doi.org/10.1016/j.icarus.2007.01.002

[16] Tonry, J. L., Denneau, L., Heinze, A. N., Stalder, B.,   Smith, K. W., Smartt, S. J., Stubbs, C. W., Weiland, H. J.,   & Rest, A. (2018). ATLAS: A high-cadence all-sky survey  system. Publications of the Astronomical Society of the  Pacific, 130(988), 064505.  https://doi.org/10.1088/1538-3873/aabadf

[17] Bellm, E. C., Kulkarni, S. R., Graham, M. J., Dekany, R.,   Smith, R. M., Riddle, R., Masci, F. J., Helou, G., Prince,   T. A., Adams, S. M., & Zolkower, J. (2019). The Zwicky

Transient Facility: System overview, performance, and first  results. Publications of the Astronomical Society of the  Pacific, 131(995), 018002.  https://doi.org/10.1088/1538-3873/aaecbe

[18] Popova, O. P., Jenniskens, P., Emel'yanenko, V., Kartashova,  A., Biryukov, E., Khaibrakhmanov, S., & Trigo-Rodriguez,   J. M. (2013). Chelyabinsk airburst, damage assessment,  meteorite recovery, and characterization. Science, 342(6162),  1069-1073. https://doi.org/10.1126/science.1242642

[19] Fast, K. (2026, February 17-18). Asteroid threats and the gap  in current surveys [Plenary remarks]. American Association  for the Advancement of Science (AAAS) Annual Meeting,

Phoenix, AZ. As reported in news coverage.[20] Grav, T., Mainzer, A. K., Masiero, J. R., Wright, E. L.,   Bauer, J. M., Cutri, R. M., Sonnett, S., & Kramer, E. A.   (2023). The Near-Earth Object Surveyor near-Earth asteroid   known-object model. The Planetary Science Journal, 4(12),  228. https://doi.org/10.3847/PSJ/ad072e

[21] Weryk, R. J., & Wainscoat, R. J. (2020). Linking isolated  tracklets to improve asteroid discovery. American  Astronomical Society Meeting Abstracts, 235, abstract  329.05. https://ui.adsabs.harvard.edu/abs/2020AAS..23532905W

[22] NASA Center for Near-Earth Object Studies. (2026).   Discovery statistics: Cumulative totals and population   by size. Jet Propulsion Laboratory, California Institute of  Technology. https://cneos.jpl.nasa.gov/stats/

[23] Minor Planet Center. (2026). Database search and provisional  designation records. International Astronomical Union,  Smithsonian Astrophysical Observatory.  https://www.minorplanetcenter.net/

[24] International Astronomical Search Collaboration. (2026).   Live campaign statistics: Preliminary, provisional, and  numbered detections. Hardin-Simmons University.

https://iasc.cosmosearch.org/

[25] Chyba, C. F., Thomas, P. J., & Zahnle, K. J. (1993). The  1908 Tunguska explosion: Atmospheric disruption of a stony   asteroid. Nature, 361(6407), 40-44.  https://doi.org/10.1038/361040a0

[26] Strasser, B. J., Tancoigne, E., Baudry, J., Piguet, S.,   Spiers, H., Luis-Fernandez Marquez, J., Kasparian, J., Grey,   F., Anderson, D., & Lintott, C. (2023). Quantifying online citizen science: Dynamics and demographics of public  participation in science. PLoS ONE, 18(11), e0293289.   https://doi.org/10.1371/journal.pone.0293289

[27] Golovich, N., Tan, T., Kuhlman, M., Wang, Q., Schneider, M. D.,   Lupton, R. H., Schein, B., et al. (2025). Survey-wide asteroid  discovery with a high-performance computing enabled non-linear  digital tracking framework. Astronomy and Computing, 53,  100987. https://doi.org/10.1016/j.ascom.2025.100987

[28] Holman, M. J., Payne, M. J., Blankley, P., Janssen, R., &   Kuindersma, S. (2018). HelioLinC: A novel approach to the  minor planet linking problem. The Astronomical Journal,

156(3), 135. https://doi.org/10.3847/1538-3881/aad69a

[29] Bolin, B. T., Denneau, L., Bressi, T. H., Hatch, P., Veres, P.,   Wainscoat, R. J., et al. (2025). The discovery and  characterization of Earth-crossing asteroid 2024 YR4. arXiv  preprint arXiv:2503.05694. https://arxiv.org/abs/2503.05694

[30] Stetzler, S., Juric, M., Bernardinelli, P. H., Bektesevic, D.,   Chandler, C. O., Connolly, A. J., Adams, F. C., et al. (2025).  An efficient shift-and-stack algorithm applied to detection

catalogs. arXiv preprint arXiv:2509.26279.  https://arxiv.org/abs/2509.26279

[31] Li, S., Geringer-Sameth, A., & Golovich, N. (2025). An  automated probabilistic asteroid prediscovery pipeline. arXiv  preprint arXiv:2510.07588. https://arxiv.org/abs/2510.07588

[32] B612 Foundation. (2025). ADAM precovery service: Open-source  implementation. Asteroid Decision Analysis and Mapping  project. https://b612.ai/opensource/precovery/

[33] United States Congress. (2005). George E. Brown, Jr. Near- Earth Object Survey Act, Section 321 of the National  Aeronautics and Space Administration Authorization Act of

2005, Public Law 109-155 (December 30, 2005). Mandated  NASA to detect, track, catalogue, and characterise the physical characteristics of 90% of near-Earth objects 140

metres or larger by 2020. https://www.congress.gov/bill/109th-congress/senate-bill/1281

[34] NSF-DOE Vera C. Rubin Observatory. (2025, June 23). Ever-  changing universe revealed in first imagery from NSF-DOE  Vera C. Rubin Observatory [Press release]. NSF, DOE, and

Association of Universities for Research in Astronomy.  https://rubinobservatory.org/news/first-imagery-rubin

[35] Cheng, A. F., Agrusa, H. F., Barbee, B. W., Meyer, A. J.,  Farnham, T. L., Raducan, S. D., & Zhang, Y. (2023).  Momentum transfer from the DART mission kinetic impact on  asteroid Dimorphos. Nature, 616(7957), 457-460.  https://doi.org/10.1038/s41586-023-05878-z