Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Research & technology· 日本語で読む

POSTECH Develops Technology to Track Outdoor 3D Motion Using Drone Flight Paths as a Reference

A POSTECH research team developed DAM3T, an outdoor 3D tracking technology using a drone and multiple cameras. It estimates camera positions and orientations from the drone’s flight path and links them to real-world coordinates using RTK-GPS. In experiments, it recorded an outdoor error of 1.21%. The method addresses camera calibration challenges in large-area measurements without requiring large on-site equipment.

Key facts

  • Jin-Tae Kim, a professor at POSTECH, and colleagues developed DAM3T, an outdoor 3D tracking technology using a drone and multiple cameras.
  • DAM3T estimates camera positions and orientations from a drone’s flight path and uses RTK-GPS positioning data to map motion to real-world coordinates, including latitude, longitude and altitude.
  • In an indoor comparison experiment in a space measuring 6 m wide × 6 m deep × 3 m high, the system recorded an error of about 2.3 cm and a normalized error of 1.15%.
  • In an outdoor experiment spanning tens of meters, the error was 1.21%; errors were below 2% both indoors and outdoors.
  • The research findings were published online in Science Advances.

What happened

TechXplore reported on November 20, 2026, that a POSTECH research team developed DAM3T, a system that tracks the 3D motion of objects across large outdoor spaces using a drone’s flight path as a reference. The team was led by Jin-Tae Kim, a professor in the Department of Mechanical Engineering, and the findings were published online in Science Advances. POSTECH issued its announcement on November 17, 2026.

Why it matters

POSTECH identifies the need for large equipment and on-site work to calibrate the positions and orientations of multiple cameras outdoors as a challenge. DAM3T uses a drone’s flight path and RTK-GPS; in reported experiments, it achieved an outdoor error of 1.21%. The method could offer an alternative for research and measurement of motion across large areas.

Analysis

The focus of this technology is not simply improving the accuracy of outdoor 3D measurements, but shifting the reference for camera calibration from structures installed on site to a flying drone. Conventional multi-camera systems require precise information about each camera’s position and orientation. POSTECH says that calibrating across large spaces using large grid-like equipment or structures has constrained measurement locations, costs and labor. DAM3T estimates camera positions from changes in the drone’s position in video and connects them to geographic coordinates via RTK-GPS. Using the same flying platform as the reference for both calibration and coordinate assignment is the method’s key innovation. However, the experimental results should be interpreted within their specific scope. The indoor result of about 2.3 cm was measured against a motion-capture system in a 6 m × 6 m × 3 m space, while the outdoor error rate of 1.21% came from an experiment spanning tens of meters. The test conditions and metrics differ, so the results do not show that the same accuracy can be achieved at all distances and in all environments. The announcement does not establish reproducibility under different outdoor areas, camera arrangements or drone flight paths, or how the system handles occluded targets. Demonstrations covered multiple drones, helium balloons used to visualize wind, and basketball players and a ball. This suggests the framework could record not only aircraft trajectories but also the motion of multiple targets and flows within a space. However, using tracking results for autonomous control also depends on factors beyond measurement accuracy, including output latency, stability during continuous operation and recovery after losing sight of a target. The published materials do not provide performance figures for these areas. The research team cited applications in atmospheric-flow monitoring, precision tracking of unmanned aircraft, autonomous control of multiple drones and sports analysis. The next points to assess are operating conditions for each application and performance when measurement results are fed into control systems. POSTECH said the research was an interdisciplinary effort following the project’s selection by the U.S. Office of Naval Research (ONR), but did not specify individual applications or deployment status for DAM3T.

Details

DAM3T uses multiple cameras to simultaneously film a drone in flight, then matches changes in its position in each video with an algorithm to automatically estimate camera positions and orientations. It does not require manual alignment. By adding centimeter-level positioning data from RTK-GPS mounted on the drone, the system maps motion to real-world coordinates, including latitude, longitude and altitude. In a comparison with indoor motion capture, the system recorded an error of about 2.3 cm and a normalized error of 1.15% in a space measuring 6 m wide × 6 m deep × 3 m high. In an outdoor experiment spanning tens of meters, the error was 1.21%; errors were below 2% both indoors and outdoors. The system tracked multiple drones flying irregularly, wind flow represented by helium balloons illuminated with LEDs, and the trajectories of outdoor basketball players and a ball.

Sources

Translated from the Japanese original: /news/postech-doroon-no-hikou-keiro-wo-kijun-ni-okugai-no-3d-dousa-ab7325. The Japanese version is authoritative if they differ.