First Triangulation Results in Search for UAP by the Galileo Project Observatories

Good to see as this is a real effort to capture images properly triangulated and we will be able to correct the resolution as well.

Still a problem as anyone looking at bigfoot images knows.  always fuzzy even when we have distance.

However just knowing two observers saw something corresponding to a real point in the sky gives us speed and a scale correction..

First Triangulation Results in Search for UAP by the Galileo Project Observatories

Avi Loeb

https://avi-loeb.medium.com/first-triangulation-results-in-search-for-uap-by-the-galileo-project-observatories-eacc4798a538

Galileo Project’s Daleks on top of Sphere in Las Vegas. (Image credit: Alex Delacroix, Galileo Project)

The fundamental limitation of all the footage provided by the Presidential Unsealing and Reporting System for UAP Encounters (PURSUE) is the lack of distance measurements to the Unidentified Anomalous Phenomena (UAPs). The projected velocity (or acceleration) of a UAP on the sky equals its angular velocity (or angular acceleration) times its unknown distance. A nearby object can cross the sky at a relatively low physical speed, whereas a very distant object would do the same only with a highly supersonic speed that cannot be achieved by birds, drones or even fighter jets. Knowing the distance is therefore crucial for resolving the nature of UAPs based on their motion. This point was recently emphasized in two reports from members of the UAP Science Advisory Council, posted here and here.

A new paper here by the Galileo Project research team, reports on its first measurements of distances to objects in the sky in search for UAP based on the method of triangulation.

The Galileo Project under my leadership aims to resolve the nature of UAP by constructing dedicated observatories for this task. Conventional astronomical observatories focus on a small portion of the sky at any given time and ignore objects that maneuver in the Earth’s atmosphere (and hence behave differently than meteors). The Galileo Project Observatories offer a novel architecture, in which a set of infrared cameras, visible-light cameras, radio sensors and acoustic microphones monitor the entire sky at all times. Our workhorse is the so-called `Dalek’ assembly of infrared or visible-light cameras.

Press enter or click to view image in full size

The infrared-Dalek is composed of an eight-camera LWIR array. Seven cameras point outwards around a circle at about 30 degrees elevation and an eighth, wider-field camera points at the zenith. (Image credit: A. Loeb et al. 2026)

The infrared-Dalek has eight FLIR LWIR Boson 640 × 512 sensors covering the wavelength band between 7.5 and 13.5 microns. Seven of these have a 50 degrees field of view and are distributed on a circle pointing outwards with a 30 degrees elevation. An eighth IR Boson camera is oriented towards the zenith. This arrangement results in a 360-degrees azimuthal and a 160-degrees elevation coverage, capable of monitoring nearly the entire sky at all times.

Press enter or click to view image in full size

The Dalek field of view in the Galileo Observatories includes the union of the eight cameras’ fields of view for a single Dalek, and the overlap between neighboring cameras. (Image credit: A. Loeb et al. 2026)

So far, three Galileo Observatories have been constructed: the first in Massachusetts, the second in Pennsylvania and the third in Nevada. The third observatory includes three observing sites in Las Vegas, to which we refer as Sphere (Latitude: 36.121 degrees, Longitude: -115.161 degrees), Hideout (Latitude: 36.072 degrees, Longitude: -115.066 degrees) and Wildhorse (Latitude: 36.057 degrees, Longitude: -115.079 degrees). Sphere lies 10.1 km from each of the other two which are 2.0 km apart, so the array is strongly asymmetric: the Hideout–Wildhorse pair constraint range less than either does against Sphere. Each of these stations is equipped with long-wave infrared (LWIR) wide-angle cameras that continuously record the skies of each site 10 degrees above the horizon. Each site operates autonomously, with onboard edge computing, GPS-disciplined timing, and automated thermal management. In one of the sites, Hideout, an Automatic Dependent Surveillance–Broadcast (ADS-B) tracks aircraft positions.

An important accomplishment in allowing the Galileo Observatories to produce reliable scientific-quality data was latency tests and calibration of the Dalek cameras. Intrinsic calibration consists of finding the focal length, optical image center and distortion coefficients that model the camera lens. Extrinsic calibration aims to find the position and orientation of the camera in space relative to an Earth reference frame. For the extrinsic calibration procedure, we used ADS-B aircraft tracks as reference sources in the absence of stars or fixed landmarks in the camera’s field-of-view, and the multi-sensor track association scheme that decides which detections belong to the same object.

One of the greatest triangulation challenges is associating the same object as seen across different cameras. When we have objects with known ground-truth positions (e.g. aircraft, the Moon or the Sun), we can first identify them in our observed tracks by projecting their 3D positions to the 2D camera pixel map and then match these to our detected tracks. Once identified, we can perform the triangulation of these objects if they share ids and overlap over time. However, this approach is restricted to identified objects, limiting the discovery of UAP. To extend the position triangulation to any object including UAP, we developed a strategy that identifies matching tracks in multiple cameras. Singling out objects seen simultaneously by different cameras poses the challenge of avoiding depth confusion, as objects with intersecting or nearly intersecting lines of sights may result in false positive matches.

Press enter or click to view image in full size

Track association by line-of-sight intersection diagram. Correctly associated detections of the same object X from different sites Sj are seen in the left panel. On the right panel, an unfavorable case where the intersecting lines of sight of different objects Oj locate a virtual object in 3D space. (Image credit: A. Loeb et al. 2026)

Our preliminary analysis involves triangulated objects during the week of May 24th to 30th, 2026. To characterize the triangulated tracks, we display the average kinematic properties of such objects in four different planes: (i) speed versus acceleration, (ii) path-length versus speed, (iii) path-length versus acceleration, and (iv) speed versus altitude. The tracks span speeds up to 298 meters per second, and accelerations may go up to 18 meters per second squared. The tracks were followed up to 27 kilometers (with a median of 3.9 kilometers) and were seen at altitudes between 3 and 12.4 kilometers.

Press enter or click to view image in full size

Kinematic distribution of the 365 objects solved from all three sites with at least 201 samples, in four projections of the same table: speed against acceleration, path length against speed, path length against acceleration, and speed against altitude. Objects linked to a ground-truth source are drawn apart from those that were not. (Image credit: A. Loeb et al. 2026)

We have identified two distinct populations which are separate from each other. One of the groups was primarily identified aircraft (284), although we also found 17 unidentified objects that have similar features. The other group had only objects that could not be linked to known aircraft (64). After inspection, we found that these objects were small clouds.

The identified aircraft show speeds between 116 and 298 meters per second, while the clouds move in agreement with wind speeds and remain under 30 meters per second. This speed difference is also reflected in the path length difference, where the aircraft are tracked for over 1 kilometer (median of 4.8 kilometers) while the clouds are rarely followed for that long (median of 0.18 kilometer). The small clouds are normally moving at 3 to 5 kilometers of altitude, while aircraft are mostly seen at the cruise altitude of around 10 kilometers (median of 10.5 kilometers).

Press enter or click to view image in full size

One object seen simultaneously across the Galileo network. Each panel is the view from one contributing camera at the same instant, with the detection visible in Sphere-camera-8 and cameras 7 at Hideout and Wildhorse. (Image credit: A. Loeb et al. 2026)

In the search for UAPs, we prioritize not assuming any global trajectory when triangulating. While this avoids imposing a given trajectory for an object, it comes at the cost of retrieving noisy tracks. Consequently, a traditional differentiation calculation of velocity and acceleration results in an overestimation of these values. Our new paper shows that applying a simple smoothing function results in a better recovery of aircraft velocities and accelerations.

Press enter or click to view image in full size

Cutouts of examples of triangulated objects, as seen from the three sites (from top to bottom, Sphere, Hideout and Wildhorse). The panels show, from left to right, unidentified objects with low speeds (below 30 meters per second), identified aircraft and unidentified objects with high speeds (above 100 meters per second). For each object, the speed and distance to the sites’ midpoint are indicated above the object’s cutouts. (Image credit: A. Loeb et al. 2026)

Altogether, the new paper — available here — demonstrates first results from a triangulation pipeline that makes no assumption about the shape of an object’s trajectory. Our triangulation array recovers three-dimensional positions of objects seen simultaneously from two or more sites by minimizing the weighted squared distance to the lines of sight at each instant, weighting each camera by its own pointing uncertainty and by its range to the target. Validated against ADS-B aircraft, this approach recovers distances within 5% for every one of the aircraft solved from three sites (66% of them being within 1%) and for 83% of the 5,626 solved from a single site pair (40% within 1%), at a median relative distance error of 1.4%; speeds are recovered within 5% for 78% and 63% of them, respectively.

The triangulated population separates in kinematic phase space without any appeal to identification. Of the 365 objects solved from three sites with at least 201 samples, 301 sit in an aircraft-like locus at a speed of 116–298 meters per second and a median altitude of 10.6 kilometers (284 of them confirmed by ADS-B and 17 unidentified) while 64 unidentified objects sit under a speed of 30 meters per second at an altitude of 3–5 kilometers and are, upon inspection, small clouds. No object falls between the two groups, so the array separates the dominant populations on dynamics alone.

Any statement about the population of unidentified objects must be made relative to a stated size, range and contrast, since the sensitivity of the array is set by apparent size and by thermal contrast of objects against the sky. No objects beyond ordinary natural and human-made ones have been identified during the analyzed week.

The triangulation results are promising. The demonstrated accuracy is what would be required to support a substantiated claim about an anomalous trajectory, were one to be found. Analysis of the recorded archive is ongoing. The best is yet to come!

***

The Galileo Project is currently designing its fourth observatory in Carl Sagan’s old residence in Ithaca, New York. The design of this observatory will be unique, aiming to resolve UAP orbs by imaging them at high resolution and taking their spectrum to infer their composition. The orb imager observatory concept was first described here.

***

We live in exciting times. The Galileo Project demonstrates that we do not need UAP whistleblowers to tell us what is in the sky. We can simply look up. The ground truth will not be dictated by viral tweets on social media but rather by evidence.