GISCI 241 Chap.11 Drone Surveys and Structure from Motion
Drone Surveys and Structure from Motion
Drone remote sensing combines flexible low-altitude acquisition with survey planning, photogrammetric geometry and geographic control. Structure from motion matches features across overlapping photographs, estimates camera positions and reconstructs three-dimensional points.
This chapter explains front and side overlap, ground sample distance, control and checkpoints, then distinguishes point clouds, orthomosaics, surface models and terrain models. It computes object height and apparent change while showing why a sharp product is not automatically geometrically accurate.
What this chapter covers
- 01
Drone platforms, sensors, flight constraints and regulation
- 02
Ground sample distance, altitude and focal length
- 03
Front overlap, side overlap and cross-flight geometry
- 04
Ground control, RTK or PPK and independent checkpoints
- 05
Feature matching, point clouds, orthomosaics and DEMs
- 06
Surface differences, validation and change uncertainty
Worked example · free
Compute height and apparent change
- +1First height equals 47.8 minus 36.2, giving 11.6 metres.
- +1Later height equals 47.1 minus 36.4, giving 10.7 metres.
- +1Apparent height change equals 10.7 minus 11.6, giving negative 0.9 metres.
- +1Check coordinate reference, vertical datum, grid alignment and surface matching between surveys.
- +1Claim physical change only if cross-survey uncertainty is smaller and unbiased.
Key terms
- Ground sample distance
- The ground length represented by an image pixel for a stated camera and flight geometry.
- Front overlap
- The shared image area between successive photographs along one flight line.
- Side overlap
- The shared image area between adjacent flight lines.
- Ground control point
- A visible surveyed target used to scale and georeference a photogrammetric model.
- Checkpoint
- An independently surveyed point reserved for evaluating model accuracy rather than fitting it.
- Orthomosaic
- A plan-view image assembled from geometrically corrected overlapping photographs.
- Structure from motion
- Reconstruction of relative camera geometry and three-dimensional structure from matched features across images.
Drone Surveys and Structure from Motion FAQ
Why does structure from motion need overlap?
The same feature must appear in several photographs from different positions so parallax can constrain camera geometry and three-dimensional location. Coverage with only one or two weak views leaves holes or unstable depth. Front, side and cross-flight overlap provide redundancy for matching and reconstruction.
Do RTK drones remove the need for control?
Precise camera positioning can reduce reliance on ground control and improve georeferencing, but independent checkpoints remain valuable for measuring accuracy. Survey geometry, camera calibration, processing and local signal conditions still matter. A model should be validated across the site and elevation range rather than trusted from equipment labels.
Is ground sample distance the same as accuracy?
No. Ground sample distance describes image sampling at the ground. Final positional and vertical accuracy also depend on overlap, control, camera calibration, feature matching, geometry, surface texture and processing. A small pixel footprint supports detail but cannot guarantee equally small coordinate error.
Why is a sharp orthomosaic not proof of survey accuracy?
Texture can look crisp even when the camera network is weak, control is clustered or the model contains broad deformation. Accuracy requires independent checkpoints, well-distributed control, sufficient overlap and inspection of point clouds and surface profiles. Ground sample distance describes image sampling; it does not by itself establish horizontal or vertical positional error.
Exam move
Sketch a flight block with front overlap, side overlap, cross-strips, control and checkpoints. Explain how each part constrains the reconstruction. Keep a product ladder from photographs to tie points, sparse cloud, dense cloud, surface model and orthomosaic, noting where scale and coordinates enter.
Practise surface minus terrain and later minus earlier calculations, then list the alignment and datum checks required before interpreting change. When you see a sharp orthomosaic, ask for checkpoint error and point-cloud structure. For project methods, memorise acquisition date, platform, camera, altitude, overlap, control, processing, coordinate reference and validation as the minimum reproducibility record.
Draw a weak flight plan with narrow overlap and parallel strips only, then redesign it with adequate front and side overlap, cross-links and coverage beyond the analysis boundary. Predict where holes, edge warping or doming may appear in each. Place control around the perimeter, interior and elevation range, and reserve separate checkpoints.
Explain why a tight control cluster can fit well locally while leaving the rest of the model poorly constrained. Trace matched features through camera alignment, sparse reconstruction, dense points, surface generation and orthomosaic creation. Mark where blur, repetitive texture, water or moving vegetation can break the chain.
Work object-height and elevation-change calculations, but do not interpret them until coordinate reference, vertical datum, grid alignment and stable-area bias are checked. Compare ground sample distance with checkpoint accuracy to reinforce that image detail is not geometric truth. For a final project audit, pair the product with a flight diagram, control map, checkpoint summary and uncertainty statement.
State the minimum change magnitude and spatial extent that can be distinguished from reconstruction error.