Point density. Digitising points off the mesh reveals the structure of the earthwork, the point density is approximately at .2m post spacing increasing to 0.1m at detail areas. A test patch at 0.05m did nothing to improve the model. As yet break-lines are absent. This preview is from about 5k points with the Z interval at o.1m:
Compared to the previous surface the difference is clear:On the left is from the February KAP flight. On the right the Sept mission, the surface disruption caused by vegetation is evident. The detail of the eroded SE outer ditch (bottom left in this view) is revealed clearly in the Feb model.
Developing the contour model. Point selection avoiding hedge and tree surfaces allows the ground surface to be developed, the quality of the surface mesh from the RX1 imagery means the true ground height is recorded to within 0.5m of the hedge lines, in places the ground height between scrub growths can be detected: it’s that good. As more points are added the contours are regenerated to check for outliers, checking against the ortho-photo confirms the selections. Point editing comes down to deleting the offending point in CAD and re-contouring.
TheoContour interpolates the points relative to a UCS with +ve values on all axes so the exported plane from PhotoScan can be dealt with by fitting a UCS to the GCPs. The flexibility of being able to try a patch, edit points and try again is something TheoContour is uniquely good at; it turns the process into a CAD job rather than a batch post processing one. The quality of the work required for this kind of site is just not going to happen with a a ‘grid it and run’ approach.
Next: more points, adding the break-lines, smoothing and annotation.