Commissioning An All-Sky Infrared Camera Array for Detection Of Airborne Objects

TL;DR

This paper documents the commissioning of an all-sky infrared camera array (the Dalek) as part of the Galileo Project's multi-modal sky observatory to monitor aerial phenomena. It presents a comprehensive calibration pipeline (intrinsic, INU removal, extrinsic with ADS-B), an object-detection/tracking framework (YOLOv5 + SORT), and a diverse suite of training/evaluation datasets, including synthetic, real, and ADS-B-derived data. The study provides baseline performance metrics: ADS-B-based acceptance around 41% and mean detection efficiency around 36% for aircraft within 10 km, and demonstrates a simulated dataset showing high tracking accuracy but with trajectory fragmentation. A toy outlier analysis using trajectory sinuosity, coupled with a likelihood-based significance test, yields an upper limit of 18,271 ambiguous outliers at 95% confidence over five months, illustrating a generalizable method for rigorous anomaly assessment. The results establish a blueprint for end-to-end instrument commissioning and lay the groundwork for a future multi-modal aerial census capable of robustly identifying truly novel phenomena.

Abstract

To date there is little publicly available scientific data on Unidentified Aerial Phenomena (UAP) whose properties and kinematics purportedly reside outside the performance envelope of known phenomena. To address this deficiency, the Galileo Project is designing, building, and commissioning a multi-modal ground-based observatory to continuously monitor the sky and conduct a rigorous long-term aerial census of all aerial phenomena, including natural and human-made. One of the key instruments is an all-sky infrared camera array using eight uncooled long-wave infrared FLIR Boson 640 cameras. Their calibration includes a novel extrinsic calibration method using airplane positions from Automatic Dependent Surveillance-Broadcast (ADS-B) data. We establish a first baseline for the system performance over five months of field operation, using a real-world dataset derived from ADS-B data, synthetic 3-D trajectories, and a hand-labelled real-world dataset. We report acceptance rates (e.g. viewable airplanes that are recorded) and detection efficiencies (e.g. recorded airplanes which are successfully detected) for a variety of weather conditions, range and aircraft size. We reconstruct $\sim$500,000 trajectories of aerial objects from this commissioning period. A toy outlier search focused on large sinuosity of the 2-D reconstructed trajectories flags about 16% of trajectories as outliers. After manual review, 144 trajectories remain ambiguous: they are likely mundane objects but cannot be elucidated at this stage of development without distance and kinematics estimation or other sensor modalities. Our observed count of ambiguous outliers combined with systematic uncertainties yields an upper limit of 18,271 outliers count for the five-month interval at a 95% confidence level. This likelihood-based method to evaluate significance is applicable to all of our future outlier searches.

Figures (33)

• Figure S1: (Left): Mechanical design drawing of the Dalek IR camera array. (Right): Photograph of the Dalek as constructed at the development site.
• Figure S2: (Left): Illustration of fields of view (FOVs) and their overlap between the eight cameras of the Dalek. The orange areas represent the FOVs of the seven cameras arranged hemispherically. The purple area shows the FOV of the zenith camera on top of the Dalek. (Right): Side view of fields of view for the hemispherical Dalek cameras. As the center of their optical axes are pointing 30$^{\circ}$ above the horizon, the bottom of the images that they capture corresponds to 10$^{\circ}$ above the horizon.
• Figure S3: Map view of a mosaic of images from the seven hemispheric cameras and the one zenith Boson IR camera, and their orientation with respect to a visible all-sky camera photograph from the Dalek's location (background, center of the image). The shaded (purple) semi-translucent overlays show the corresponding treeline masks that are used in post-processing to ignore all but the sky area of the images. All camera frames are taken from a video recording from 7 May 2024 except for camera 1 (3 April 2024) and camera 7 (10 May 2024).
• Figure S4: Stacked area plot showing the evolution over time of the sum of all cameras' recording efficiencies, defined as recording duration per camera per day divided by expected duration based on recording schedule, which varies per camera. If all cameras were recording according to the schedule all the time, the summed efficiencies should add up to 8. The few data points above 8 are due to manual enabling of the recording, for testing purposes. This timeline goes from November 2023 to May 2024. Some cameras, such as cameras 6 and 7, show a drastic improvement over time.
• Figure S5: Metal chessboard used for intrinsic calibration of the FLIR Boson 640 cameras. (Left): Boson camera image. (Right): Dark-painted metal cutout chessboard on dark-painted metal base plate.