03 · SCHOLARLY
304 FILES·LAST 5D AGO
← Scholarly / Papers
SCHOLARLY / PAPER

Detection of Moving Objects in Earth Observation Satellite Images

Eric Keto · Wesley Andres Watters

Journal of Astronomical Instrumentation · 2023

Keto & Watters (2023) show that Planet Labs SuperDove push-broom satellite imagery encodes millisecond-resolution motion data across eight spectral bands, enabling detection and velocity measurement of aircraft, cars, and boats, and frames the technique as a Galileo Project search modality for anomalous aerial objects.

SharePostReddit

Brief

The paper demonstrates two motion signatures latent in multi-spectral push-broom imagery: displaced color ghosts in RGB composites and positive-negative pairs in differenced adjacent spectral band images. Applied to a single Planet Labs scene over San Diego (May 30, 2022, satellite 241e), the authors recover a ~0.39 s inter-band time delay from orbital ephemeris alone and measure an airplane's ground speed at 38.5 ± 5.3 m/s (86.4 ± 11.8 mph). The minimum detectable ground speed is ~20 m/s, set by the 3 m resampled pixel size, and an altitude-speed ambiguity of less than 10% applies to any object below the stratosphere.

Metadata

Category
Hub & Overview
Venue
Journal of Astronomical Instrumentation
Type
Peer-reviewed
Year
2023
Authors
Eric Keto, Wesley Andres Watters
Access
Open access
Length
12.0 M
Programs
Galileo Project
Instruments
Planet Labs SuperDove satellites
Data sources
Planet Labs archive (basic analytic images), Planet Labs daily ephemeris (TLE / mean motion)
Tags
UAP-detection, remote-sensing, image-processing, technosignature, aerial-surveillance, SETI

Key points

  • Two distinct motion signatures exist in SuperDove imagery: (1) three displaced color ghosts in RGB composites, and (2) positive-negative pairs in images differenced between spectrally adjacent bands.p.3
  • The inter-band time delay is derived purely from published orbital parameters (mean motion: 15.15 orbits/day) and sensor geometry (663-pixel color strip width); for satellite 241e the value is 0.393562 s.p.5
  • Minimum detectable ground speed is ~20 m/s, set by the 3 m ground sample distance of the orthorectified archive images.p.1
  • Airplane at San Diego International Airport measured at 38.5 ± 5.3 m/s (86.4 ± 11.8 mph), well below typical Boeing 737 takeoff speed of 150 mph, consistent with taxiing.p.7
  • Planet Labs archive images are mosaics of sequential exposures and carry no unique per-pixel timestamps; timing must be reconstructed indirectly from camera frame rate (0.17 s) and orbital velocity.p.3
  • Stationary elevated objects (e.g., clouds) produce parallax-induced color displacement that mimics a motion signature because orthorectification is referenced to ground level.p.8
  • Altitude-speed ambiguity is bounded: the stratosphere tops out at ~50 km versus a ~500 km orbital altitude, yielding less than 10% velocity uncertainty for any atmospheric object.p.8
  • The method is positioned as a Galileo Project search modality to automatically flag objects whose velocity, acceleration, size, or shape fall outside the envelope of natural phenomena and known vehicles.p.9

Verbatim

  • Moving objects have characteristic signatures in multi-spectral images made by Earth observation satellites that use push broom scanning.
    p.1
  • The minimum ( ∼ 20 ms − 1 ) detectable ground speeds is set by the spatial resolution.
    p.1
  • The average speed of the airplane in the time between these images is 38 . 5 ± 5 . 3 m/s or 86 . 4 ± 11 . 8 mph.
    p.7
  • For objects within the Earth's atmosphere, such as those that may be detected in the Planet Labs images, unusual flight patterns have been suggested as an indicator of extraterrestrial origin (Office of the Director of National Intelligence 2021).
    p.9
  • This research is a first step to develop a method to observe large areas of the Earth, automatically recognize moving objects, and select those whose velocity, acceleration, size or shape fall outside those expected for natural phenomena or common vehicles or projectiles (Watters et al. 2023).
    p.9
  • This research is conducted as part of the Galileo Project (Loeb & Laukien 2023) at Harvard University whose goal is to collect scientific quality data that may be useful in the search for objects of extraterrestrial origin
    p.9

Most interesting

  • A single published orbital parameter, mean motion in orbits per day, is sufficient to reconstruct millisecond inter-band timing without access to Planet Labs' internal image-processing pipeline.
  • The SuperDove sensor's eight spectral bands effectively convert a single still image into a ~0.39 s temporal sequence, yielding a rudimentary eight-frame 'video' at no additional observation cost.
  • The camera frame interval of 0.17 s causes a 1.2 km ground shift between exposures; a moving object near a strip boundary introduces a discrete velocity ambiguity of factor ~0.5 or 1.5, resolvable by minimizing apparent acceleration across the band sequence.
  • The 2021 ODNI Preliminary Assessment on UAP is cited directly as scientific motivation for selecting objects with anomalous flight patterns, one of the few peer-reviewed papers to formally cite a government intelligence document as a search criterion.
  • Clouds at altitude routinely appear to move in Planet Labs imagery due to parallax, providing a built-in calibration case for the altitude-speed ambiguity described in equation 3.
  • The scene analyzed covers 40.0 × 29.7 km² at 3 m resolution (13,317 × 9,578 pixels), making wide-area automated search computationally tractable with standard difference-imaging techniques originally developed for supernova detection in astronomy.

Cross-references