ONC: Hayabusa2 Optical Navigation & Science

• Optical Navigation Camera (ONC) is a three-camera system on Hayabusa2 that delivers both navigational and multispectral scientific imaging for asteroid Ryugu.
• Advanced calibration and image registration techniques ensure sub-pixel accuracy and radiometric precision, critical for mapping geomorphology and detecting subtle spectral features.
• ONC-T’s dual role in cross-calibration, sample-return analysis, and even exoplanet transit detection underscores its versatile contribution to small-body science and space photometry.

Searching arXiv for recent and core ONC papers to ground the article. arXiv search: Hayabusa2 Optical Navigation Camera ONC-T image registration Ryugu The Optical Navigation Camera (ONC) is the three-camera imaging system aboard JAXA’s Hayabusa2 spacecraft, comprising the telescopic ONC-T and the wide-angle ONC-W1 and ONC-W2. Although conceived as navigation hardware, the ONC became a dual-use system for optical navigation and quantitative remote sensing of asteroid 162173 Ryugu. In particular, ONC-T, a telescopic CCD framing camera with seven filter bands spanning the ultraviolet, visible, and near-infrared, provided the multispectral dataset used to map Ryugu’s spectral distribution, while the wide-angle cameras supported descent, touchdown, hazard monitoring, and close-proximity operations (Tatsumi et al., 2018, Kouyama et al., 2021, Hirata et al., 13 Feb 2026).

1. System configuration and instrument roles

Hayabusa2’s ONC system consisted of three cameras with distinct geometries and operational functions (Tatsumi et al., 2018, Hirata et al., 13 Feb 2026).

Camera	Configuration	Principal role
ONC-T	Telescopic, narrow-angle, multi-band CCD camera	Optical navigation and scientific imaging
ONC-W1	Wide-angle, nadir-looking camera	Descent, touchdown, hazard monitoring
ONC-W2	Wide-angle, slant-viewing camera	Close-proximity support and complementary imaging

ONC-T was the primary scientific camera. It had seven band-pass filters spanning approximately 0.4–0.95 μm, and in the calibration paper the filter set is given as ul (0.40 μm), b (0.48 μm), v (0.55 μm), Na (0.59 μm), w (0.70 μm), x (0.86 μm), p (0.95 μm), and wide (Tatsumi et al., 2018). In the exoplanet-photometry study, ONC-T is further described as a multi-band, shutter-less CCD camera with focal length 120.5 mm, effective aperture diameter 15 mm, a 1024 × 1024 pixel CCD, 100% fill factor, and a panchromatic filter covering 400–950 nm (Yumoto et al., 16 Oct 2025).

A recurrent misconception is that the ONC was merely a navigation subsystem. The mission record described in the calibration, registration, and mapping studies shows otherwise: ONC-T provided spectral mapping and compositional constraints, while ONC-W1 and ONC-W2 were indispensable during low-altitude operations when wide field of view and close-range geometry dominated operational requirements (Tatsumi et al., 2018, Hirata et al., 13 Feb 2026). The statement in the mapping study that ONC was “crucial for mapping geomorphology and composition such as hydrated minerals during navigation and scientific observation” captures this dual role directly (Hirata et al., 13 Feb 2026).

2. Optical, spectral, and radiometric characteristics

The scientific utility of ONC depended on inflight calibration. During the 3.5-year cruise after Hayabusa2’s launch on 3 December 2014, the ONC team conducted calibrations of read-out smear, electronic-interference noise, bias, dark current, hot pixels, sensitivity, linearity, flat-field, and stray light for the full camera suite (Tatsumi et al., 2018). The same work states that the updated ONC-T absolute radiometric calibration contains less than 1.8% error for the ul-, b-, v-, Na-, w-, and x-bands based on star calibration observations and ~5% for the p-band based on lunar calibration observations (Tatsumi et al., 2018).

The calibrated conversion from raw ONC-T data to physical radiance is written as (Tatsumi et al., 2018)

$$F = L^{-1}\!\left(I_{\mathrm{raw}} - I_{\mathrm{bias}(T_{\mathrm{CCD},T}, T_{\mathrm{ELE},T}, T_{\mathrm{AE},T})} - I_{\mathrm{smear}(I_{\mathrm{raw}})} - I_{\mathrm{dark}(t,T_{\mathrm{CCD},T})} - I_{\mathrm{st}} \right) \big/ \left(f_{r,n}\, S_n(T_{\mathrm{CCD},T})\, t\right).$$

This expression formalizes the sequence emphasized throughout the calibration study: subtract instrumental artifacts, apply linearity correction, then divide by flat-field, sensitivity, and exposure time. The same paper reports that ONC-T sensitivity at the reference CCD temperature of $-30^\circ$C was 439.1 ± 2.2 for ul, 969.3 ± 7.9 for b, 1175.0 ± 10.0 for v, 546.9 ± 2.0 for Na, 1515.0 ± 19.3 for w, 1499.8 ± 24.6 for x, and 1033.0 ± 71.1 for p in $$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$, with 961.2 ± 28.8 adopted for p-band pipeline use from lunar calibration (Tatsumi et al., 2018).

Several detector effects were sufficiently large to require explicit modeling during science operations. The ONC-T bias model is given as

$I_{\mathrm{bias}(T_{\mathrm{CCD}, T_{\mathrm{ELE}})} = 320.66 + 0.652\,T_{\mathrm{CCD}} - 0.953\,T_{\mathrm{ELE}},$

with an additional AE-temperature correction

$B(T_{\mathrm{AE}})=0.987-0.00251\,T_{\mathrm{AE}}.$

The dark-current model is

$$I_{\mathrm{dark}} = t\,\exp(0.10\,T_{\mathrm{CCD}} + 0.52)\;[\mathrm{DN}],$$

yielding <0.09 DN/s at $-30^\circ$C and about 20 DN/s at $20^\circ$C (Tatsumi et al., 2018). A pixel with dark current >30 DN/s is classified as a hot pixel, and under touchdown conditions as much as ~1% of the FOV may be affected by hot pixels exceeding 100 DN/s (Tatsumi et al., 2018).

Linearity was also quantified. The inflight relation remained good to ~3100 DN with <1% deviation, while nonlinearity reached 10–13% at about 3600 DN; after applying the published correction polynomial, linearity errors were reduced to <0.1% (Tatsumi et al., 2018). Flat-field behavior was band dependent: ONC-T v-band sensitivity was uniform within ~2%, whereas ul-band uniformity was around ~3.9%, with corner regions 6–9% lower than average (Tatsumi et al., 2018). This matters because Ryugu is spectrally subtle and very dark; the laboratory-comparison study gives the ONC-T global average reflectance factor at 550 nm as about 1.88 ± 0.17%, underscoring why small calibration errors can become scientifically consequential (Cho et al., 2021).

3. Navigation, descent operations, and the Ryugu observation record

The ONC supported both optical navigation and scientific observation over the full Ryugu campaign. The geometry-refinement study states that ONC collected more than 8,300 images of Ryugu, revealing its spinning-top shape and boulder-covered surface (Hirata et al., 13 Feb 2026). Those images spanned the full mission timeline, from global observations at the home position to very high-resolution descent images and touchdown sequences, including the artificial crater experiment (Hirata et al., 13 Feb 2026).

The division of labor among the cameras reflected spacecraft proximity and required scene context. ONC-T provided the primary multispectral record for color/compositional variations, hydrated-mineral searches, and global scientific interpretation (Hirata et al., 13 Feb 2026). ONC-W1 and ONC-W2 were especially important during descent and touchdown operations because of their wide fields of view; the calibration paper adds that ONC-T is out of focus at very short distances, which further explains the operational importance of the wide-angle cameras during touchdowns, gravity measurements, and payload deployments (Tatsumi et al., 2018).

The mission-specific significance of ONC-T lies in its sequential multi-band imaging strategy. Because multi-band observations were acquired one filter at a time rather than simultaneously, Ryugu’s rotation and small changes in spacecraft position and attitude caused the same surface feature to fall on slightly different pixel locations in different bands (Kouyama et al., 2021). This characteristic produced a direct coupling between acquisition geometry and downstream spectral analysis. The registration paper explicitly notes that even a shift of one or a few pixels can create false spectral features in band ratios and color composites, especially in textured regions (Kouyama et al., 2021).

The calibration paper also translated instrument performance into science-operational limits. It estimates ~1.6% uncertainty in the ONC-T 0.7 μm hydration-band depth metric

$d_{0.7} = 1 - \frac{1.6R_p + 1.5R_x}{3.1R_w},$

and concludes that typical serpentine-like absorptions of 3–4% should be detectable at SNR ~2 (Tatsumi et al., 2018). For sodium-emission searches, it reports that a single v+Na image set can detect sodium atmospheres of several 10s of kR, while integrating 100 image sets improves detectability to several 100 R (Tatsumi et al., 2018). At high CCD temperatures near 20°C, ONC-T SNR can drop to ~150, yet the camera still permits detection of about 3% radiance variations (Tatsumi et al., 2018). These are mission-use figures rather than generic optical-camera specifications, and they tie the ONC directly to Ryugu science and operations.

4. Image registration, bundle adjustment, and map products

A substantial fraction of ONC science depended not only on instrument calibration but also on geometric refinement. For ONC-T multi-band observations, image co-registration was required because the multi-band sequence was not simultaneous (Kouyama et al., 2021). The registration workflow combined SURF-based rough matching, RANSAC outlier removal, coarse-to-fine template matching repeated three times, and a hierarchical transform model consisting of a global affine transform followed by local affine transforms based on nearby optical-flow vectors (Kouyama et al., 2021). Wrong motion vectors were rejected using neighboring-vector consistency and correlation quality, and final resampling was performed once with bilinear interpolation to reduce blurring (Kouyama et al., 2021).

The paper validated the method using both real ONC-T images and an artificial displacement test. In that test the image was shifted by 0.5 pixels in $x$ and by a sinusoidal displacement in $-30^\circ$0,

$-30^\circ$1

and the recovered shift was about $-30^\circ$2 pixels in $-30^\circ$3, with similar $-30^\circ$4-pixel error in $-30^\circ$5 (Kouyama et al., 2021). The overall registration accuracy reached the order of 0.1 pixels, which the authors judged sufficient for many Ryugu spectral-mapping applications (Kouyama et al., 2021). The same study illustrated the effect with Earth-image validation using the normalized difference vegetation index,

$-30^\circ$6

showing that artificial edge enhancement disappeared after registration (Kouyama et al., 2021). In the Ryugu context, this removed false color artifacts and stabilized band-ratio interpretation.

At a larger geometric scale, the 2026 mapping study addressed the fact that most high-resolution descent and touchdown images lacked precise location and camera pose metadata (Hirata et al., 13 Feb 2026). It refined geometry for 994 ONC images, while about 7,600 images already had precise geometry from an earlier SPC-derived bundle (Hirata et al., 13 Feb 2026). All images from 11 descent operations were targeted, including MINERVA-II1 deployment, MASCOT deployment, TD1 and TD2 rehearsals, TD1 and TD2 touch-down sequences, and SCI crater-search/descent observations (Hirata et al., 13 Feb 2026). The highest-resolution images reached roughly 0.35 mm/pixel at the lowest altitudes (Hirata et al., 13 Feb 2026).

The bundle-adjustment procedure was a nonlinear least-squares refinement of control-point coordinates and camera pose. The study implemented a custom iterative tool using SciPy, NumPy, OpenCV, and Open3D, solved with Levenberg–Marquardt, and generated more than 10,000 control points (Hirata et al., 13 Feb 2026). The published objective is the standard reprojection-error minimization

$-30^\circ$7

For descent images, the control-point 3D coordinates were constrained to the Ryugu shape model; L2b images were used for geometry estimation, with internal distortion correction applied to image coordinates because L2d images had cropped edge pixels (Hirata et al., 13 Feb 2026). For ONC-T, the study assumed a fixed focal length of 120.5 mm corresponding to the v-band, with small inter-filter differences absorbed into the surface-distance term (Hirata et al., 13 Feb 2026).

The resulting products included GeoTIFF map-projected files, spatial-resolution maps, emission-angle maps, and solar-incidence-angle maps, all with embedded geographic metadata suitable for QGIS or ArcGIS (Hirata et al., 13 Feb 2026). Global and regional mosaics were then assembled, including western-hemisphere mosaics before and after the SCI experiment, equatorial mosaics, polar maps in polar azimuthal equidistant projection, and local mosaics for TD1, TD2, MINERVA, and MASCOT (Hirata et al., 13 Feb 2026). Before mosaicking, pixels with emission angles > 65° were excluded, non-overlapping pixels from synchronous seven-band images were removed, and brightness/contrast of L2d images was normalized to match reference data (Hirata et al., 13 Feb 2026). These products, together with refined geometry and metadata, were released publicly at https://doi.org/10.7910/DVN/WW3IH0 under CC BY 4.0 (Hirata et al., 13 Feb 2026).

5. ONC-T as a reference instrument for cross-calibration and sample-return interpretation

The ONC-T dataset became a radiometric reference not only within Hayabusa2 but also in cross-mission and laboratory contexts. In the Ryugu–Bennu comparison study, ONC-T served as the Ryugu reference instrument against which OSIRIS-REx/MapCam data were scaled (Yumoto et al., 2023). The central problem was an imager-to-imager systematic bias of up to 15% caused by differences in radiometric calibration targets and solar irradiance models (Yumoto et al., 2023). ONC-T had been calibrated using standard stars, whereas MapCam used the Moon via ROLO-based lunar calibration (Yumoto et al., 2023).

The paper expresses observed reflectance in band $-30^\circ$8 as

$-30^\circ$9

and derives a total cross-calibration factor

$$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$0

Using the Moon as a common standard, the authors found that pre-cross-calibrated Bennu/MapCam reflectance must be scaled upward by 13.3 ± 1.6% at b, 13.2 ± 1.5% at v, 13.6 ± 1.7% at w, and 14.8 ± 1.8% at x, while Ryugu ONC-T reflectance was kept unchanged (Yumoto et al., 2023). After correction, Ryugu and Bennu reflectance could be compared with <2% accuracy, and Bennu’s geometric albedo became consistent with ground-based telescopes and OVIRS (Yumoto et al., 2023). This establishes ONC-T not merely as a mission camera but as a stable photometric reference within small-body comparative planetology.

A second continuity role appears in returned-sample analysis. A laboratory multispectral stereo-camera system was intentionally designed to be comparable to ONC-T, using six ONC-T-like bands: 390 nm (ul), 475 nm (b), 550 nm (v), 590 nm (Na), 700 nm (w), and 850 nm (x) (Cho et al., 2021). The 950-nm p-band was omitted because it was not used often for nominal ONC-T observations and the CMOS detector had low quantum efficiency there (Cho et al., 2021). The system produced 4096 × 2160 images at 1.93 μm/pixel over a 7.9 × 4.2 mm field of view, and validation showed ~3% relative reflectance spectral error and 5% 3D-model error (Cho et al., 2021).

This laboratory bridge allowed direct comparison between ONC-T remote sensing and Ryugu returned grains. Measurements of samples from dishes A3 and C1 showed average spectra that were flat and consistent with the global averaged spectrum of Ryugu, while the 550-nm (v-band) reflectance of the returned grains was 2.4% on average, higher than that of the global averaged spectrum of Ryugu observed with ONC-T (Cho et al., 2021). The authors suggest that this apparent offset could reflect greater specular reflectance, reflective opaque minerals, or differences in microtexture and porosity (Cho et al., 2021). This suggests that ONC-T’s global-scale radiometry is sufficiently stable to support laboratory-to-spacecraft representativeness tests rather than only qualitative visual comparison.

6. Repurposing ONC-T beyond Ryugu and broader implications

The clearest demonstration that ONC-T exceeded its nominal navigation role is the 2025 exoplanet-photometry study. During Hayabusa2’s cruise phase, the team used ONC-T to observe transits of WASP-189 b and MASCARA-1 b (Yumoto et al., 16 Oct 2025). They collected data for ten and four events, respectively, with roughly 21 hours of continuous coverage per event (Yumoto et al., 16 Oct 2025). The data reduction included corrections for bias, dark current, hot pixels, and flat fielding, followed by aperture photometry and detrending with pixel-level decorrelation plus linear and exponential time-dependent terms to account for pointing drift and temperature evolution (Yumoto et al., 16 Oct 2025).

The corrected image was written as

$$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$1

and the light curve was modeled schematically as

$$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$2

with

$$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$3

The transit signal-to-noise ratio was defined as

$$(\mathrm{DN/s})/(\mathrm{W\,m^{-2}\,\mu m^{-1}\,sr^{-1}})$$4

The principal result was that ONC-T detected the transit signal with SNR 13 for WASP-189 b and 8 for MASCARA-1 b for each event, improving to 40 and 16 after stacking all events (Yumoto et al., 16 Oct 2025). Single-event transit mid-times were measured with precision of about 6 minutes, consistent with TESS to within 2 minutes, and the planet-to-star radius ratio was determined with absolute precision 0.004 and agreement with TESS to within 0.002 (Yumoto et al., 16 Oct 2025). The study also reports a 4 sigma discrepancy between the updated orbital period of MASCARA-1 b and previously reported values (Yumoto et al., 16 Oct 2025).

These detections were not merely threshold events. A blind Transit Least Squares search recovered the correct orbital periods, with signal detection efficiencies of 9.1 for WASP-189 b and 6.2 for MASCARA-1 b (Yumoto et al., 16 Oct 2025). The paper concludes that ONC-T set a new record for the smallest-aperture instrument to detect an exoplanet transit from space (Yumoto et al., 16 Oct 2025). A plausible implication is that the engineering constraints that shaped ONC-T—small aperture, compact optical train, and navigation-oriented stabilization environment—did not preclude quantitatively useful time-domain photometry when acquisition strategy and detrending were adapted to the instrument.

Taken together, the Ryugu mapping papers, the cross-calibration study, the sample-return comparison system, and the cruise-phase transit analysis show that ONC, and ONC-T in particular, should be understood as a mission-integrated optical system whose significance lies in calibration, geometry, and operational versatility as much as in detector hardware. Within Hayabusa2 it enabled navigation, compositional mapping, geomorphologic comparison, and publicly reusable GIS products; beyond Ryugu it served as a radiometric reference for asteroid intercomparison and as a demonstrated space-based photometer at the 15-mm aperture scale (Hirata et al., 13 Feb 2026, Yumoto et al., 2023, Yumoto et al., 16 Oct 2025).