Videogrammetry reconstructs a scene from images extracted from a video. With ROLAVICam, AutoTags and ROLAVIGeoStudio, the task is to preserve the link between each frame, its time and the available references for processing in Agisoft.
One video, many connected images
Instead of taking individual photographs, you record a continuous route. You then select sharp frames with enough overlap to reconstruct the object. Moving around the scene provides different views; rotating from one spot does not provide the same geometry.
Videogrammetry in action
Watch the ROLAVICam demonstration that accompanies this workflow.
The videogrammetry workflow
The video provides images; GNSS and AutoTags provide references. Camera + IMU support (VIO) can help maintain relative continuity when RTK weakens, provided the capture records that data.
*The dashed branch applies only if the version and capture include recorded VIO. The IMU alone does not replace RTK. This diagram does not imply the entire integration is available in every version or guarantee a particular accuracy.
From field to office: keep the original MP4, its records and AutoTag detections. GeoStudio links frames by ID and time, prepares controls and enables geotagging when valid camera references exist.
For better accuracy: use surveyed, well-distributed AutoTags as GCPs. Their observations support the adjustment; accuracy is checked against other independent points.
Prepare and record in 4 steps
- Place and survey the AutoTags. Fix them to stable surfaces. With Take Point / Mark, record each tag center and retain the same ID for processing.
- Check the equipment. Check the GNSS connection, solution quality and phone mount. Record the antenna-to-camera geometry when camera positions will be used.
- Record slowly in good light. Keep the framing steady, avoid sudden zoom changes and observe tags from several positions. Aim for continuous coverage without hidden areas.
- Keep the originals. Do not trim or change the speed of the video before matching it to the records. An edit can break the timing correspondence.
A frame needs its actual time
MP4 frames and GNSS observations may have different rates. PTS identifies a frame's time within the video; associating it with GNSS requires a verified timing relationship.
For example, a 5 Hz GNSS log provides one observation every 0.2 s. Not every intermediate image has its own observed position. Select frames for sharpness and coverage, and relate the data using actual times.
An overlap target near 90% between selected images can provide guidance, but a fixed extraction interval cannot guarantee it: movement speed and distance to the object also matter.
Combine, geotag and prepare GCPs
In the compatible ROLAVIGeoStudio video workflow, bring together the MP4, GNSS records and AutoTag files. Review detections and extract the images corresponding to recorded times, keeping their identifiers.
Camera geotagging is appropriate only with valid positions, synchronized timing and known mounting geometry. An AutoTag center coordinate is a scene control point; it must not be copied as the camera position.
If usable camera positions are absent, prepare the images with their GCPs. Agisoft can estimate cameras during alignment and use those surveyed points to reference the model. The AutoTag workflow must not fill GNSS gaps with invented coordinates.
Check that every GCP retains its ID, coordinates, height reference and marked center in the images. Take the project to Agisoft Metashape Professional for alignment, adjustment and 3D model generation.
Check before delivering
Review aligned cameras, gaps and duplicated surfaces. Compare known distances and errors at surveyed points that were not used as adjustment controls.
A visually attractive model does not establish accuracy by itself. Quality depends on the video, control distribution and consistency of references and timing.
If you prefer individual photographs, see the article on terrestrial photogrammetry with ROLAVICam and AutoTags.
Technical references: Google · Visual-inertial tracking · Agisoft · Video import and processing · Control and check points.