Drone Photogrammetry for a Housing-Scheme Base Map
A GIS unit is asked to produce a 1:1,000 topographic base map and earthwork volumes for an 80 ha housing scheme in Islamabad. A fixed-wing drone flies at 120 m above ground level with 80% forward and 60% side overlap. The team lays 12 ground control points, but 9 are on rooftops and 3 are at the site boundary; no independent checkpoints are used. The orthomosaic looks sharp. After processing, the horizontal differences at three independent RTK checkpoints are 0.28 m, 0.31 m, and 0.22 m; vertical differences are +0.55 m, +0.61 m, and +0.48 m. The client wants to use the surface to compute cut-and-fill and to set out plot corners. The project specification requires 0.10 m horizontal and 0.15 m vertical accuracy at 95% confidence. The engineer must decide whether to accept the deliverable and what to recommend.
1. Interpret the checkpoint results against the specification and identify the likely causes of error.
Read the worked answer
Horizontal RMSE is sqrt((0.28² + 0.31² + 0.22²)/3) = 0.27 m. Vertical RMSE is sqrt((0.55² + 0.61² + 0.48²)/3) = 0.55 m. At 95% confidence with only three checkpoints, a conservative multiplier is needed; even 1.73 times RMSE gives about 0.47 m horizontal and 0.95 m vertical, far beyond 0.10 m and 0.15 m. The near-constant positive vertical offset of 0.48–0.61 m suggests a datum or geoid problem, or biased control heights, rather than random noise. Nine ground control points on rooftops are not at ground level and bias the bundle adjustment; boundary-only control gives poor interior geometry; and the absence of checkpoints means the adjustment was never independently validated. A sharp orthomosaic is a visual product, not an accuracy statement.
2. Recommend a revised field and processing design that could meet the specification.
Read the worked answer
Place ground control points on stable ground and distribute them across the block, including interior points, with at least 8–10 well-spaced points. Measure them by RTK or static GNSS tied to validated control, and measure their heights by level or total station if vertical accuracy matters. Collect at least 20 independent checkpoints that are not used in the adjustment. Fly lower or with higher overlap, verify camera calibration, and process with the correct geoid or orthometric height model. Report RMSE and 95% accuracy from the checkpoints, not from the adjustment residuals. For 0.10 m horizontal accuracy, 120 m above ground level with a fixed-wing drone may be marginal; a lower altitude or RTK/PPK drone should be considered.
3. Advise on the appropriate use of the current dataset while a reflight is organised.
Read the worked answer
The current dataset is acceptable for reconnaissance, visualisation, and rough planning only. It should not be used for setting out plot corners, earthwork payment, or design surfaces. Volumes should not be issued: a vertical bias of about half a metre over 80 ha could change cut-and-fill quantities substantially. If the data must be used, label them as indicative, state the checkpoint accuracy, and keep them separate from contractual deliverables. Recommend a reflight before acceptance and document the non-conformance against the 0.10 m and 0.15 m specification.