Database Reference
In-Depth Information
UAV
photogrammetry
in
PostGIS
-
orthorectification
In the previous recipe, we explored the initial steps in photogrammetric processing,
withPostGISasourstorageendpointforthepointcloudderivedfromaerialimagery.
Thenextstepistocreateimageryfromthispointcloudintheplanview,thatis,in2D
map coordinates.
Toderivetheplanvieworthophotographyfromthepointcloud,weneedtodoseveral
things.First,weneedamethodtoconvertthepointcloudtoa2Darealrepresenta-
tion.Wecoulduseformalinterpolation;but,forthesakeofsimplicity,wewilldothis
usingVoronoipolygons,aspace-fillingapproachthatallowsustoconvertourpoints
to polygons. Next, we need to attribute those polygons with the colors derived from
the original imagery. Finally, we need to render those polygons to raster.
This is not a complete orthorectification approach and fails badly where our point
cloudisleastdense;but,intheabsenceofadrapefunctionwithaPostGISrasterfor
draping our imagery over a digital surfacemodel, this is a good first approximation.
Getting ready
Wewillusethepointcloudfromthepreviousrecipeforthisexercise.Recallthatwe
have a point cloud with x, y, and z values, normal values nx, ny, and nz, and 24-bit
color values separated into three integer columns: red, green, and blue.
Theprocessingwewilldowiththispointcloudwillberelativelycompute-timeintens-
ive, so let us start off with a subset of our data while we work our way through the
steps.
CREATE TABLE chp07.uas_subset AS
SELECT * FROM chp07.uas
ORDER BY RANDOM()
LIMIT 5000;
Moving temporarily from more than 240,000 records to 500.
Search WWH ::




Custom Search