Geo-referenced Images#

Definition#

A geo-referenced image is an image whose pixels are tied to real-world coordinates through a known sensor pose (position and orientation) and camera model. Each pixel can be mapped to a ground location in a chosen spatial reference system (e.g., WGS84, UTM), enabling measurement, overlay with GIS layers, and integration into mapping pipelines.

Where They Come From#

  • Inertial Navigation Systems (INS: IMU + GNSS): Cameras paired with IMU/GNSS log position/attitude at capture time; the exterior orientation is attached to each frame.

  • Bundle blocks / photogrammetry rigs: Overlapping images from calibrated cameras are processed in a bundle adjustment to solve exterior orientations and sparse 3D structure.

  • Drones / aerial vehicles: Flight controllers provide GNSS + IMU; trigger times are synced to images; optional onboard RTK/PPK improves accuracy.

  • Crew-flown aircraft: Similar to drones but often with multi-camera nadir/oblique setups and higher-grade INS.

  • Computer vision pipelines: SLAM/VO systems estimate camera trajectories; when anchored with GNSS or surveyed control, the trajectory becomes geo-referenced.

Information Needed for a Fully Geo-referenced Image#

To compute a reliable mapping from pixels to ground coordinates you need:

  • Exterior orientation: camera position (X, Y, Z) and attitude (roll, pitch, yaw) in a known datum/CRS at the exposure time.

  • Interior orientation (camera intrinsics): focal length, principal point, lens distortion, sensor format; ideally from a calibrated camera model.

  • Timing alignment: precise synchronization between shutter time and navigation data (GNSS/IMU or SLAM trajectory).

  • Coordinate reference definition: horizontal and vertical datums, projection (e.g., EPSG code), and any geoid model used for orthometric heights.

  • Metadata linkage: image filename <-> navigation record, plus quality flags (fix type, PDOP, lever-arm applied?).

Camera Calibration#

  • Pre-calibration (lab/field): The camera is calibrated before the mission using targets or checkerboards to solve intrinsics and lens distortion; results are stable if focus and temperature are controlled.

  • On-the-job self-calibration: Calibration parameters are estimated during bundle adjustment/SLAM from the mission imagery itself; reduces bias from changing focus/temperature but requires good network geometry (overlap, roll/yaw variation, crossed flight lines).

  • Hybrid: Start from a lab prior and allow limited adjustment in processing to absorb small in-flight changes.

Lever Arms and Boresight#

  • Lever arm: 3D offset between the GNSS/IMU reference point and the camera center. Must be measured (survey) and applied.

  • Boresight: Rotation between IMU body frame and camera frame; estimated via calibration flights or target campaigns.

Accuracy Considerations#

  • GNSS quality: Single-frequency vs RTK/PPK vs PPP; update rate; multipath environment.

  • IMU grade: MEMS vs tactical; affects short-term attitude accuracy between GNSS updates.

  • Flight geometry: Forward/side overlap, cross strips, varying heights/attitudes improve bundle strength.

  • Timing: Sub-millisecond trigger tagging reduces position drift in fast motion.

  • Ground control: Well-distributed GCPs/Check points anchor the solution and validate absolute accuracy.

  • DEM/height model: Needed for orthorectification and footprint projection.

Typical Outputs#

  • Geo-tagged images (EXIF with GPS tags plus custom fields for attitude).

  • Exterior orientation tables (CSV/JSON) per image.

  • Orthophotos.

Practical Tips#

  • Keep a metadata manifest: image id, timestamp, GNSS fix type, PDOP, attitude, lever-arm applied?, boresight version.

  • Record firmware and calibration versions with the dataset.