Create_Orthophotos_from_UAV_images
Generating a 3-D point cloud + Orthophotography from UAV digital photography with VSfM-CMPMVS-GDAL2Tile
Step 1: Open Visual-SfM
Select images from a UAV flight and load into V-SfM
Run SIFT
Run Sparse Point Cloud Generation and Bundle Adjustment
You will see the camera poses begin to populate the model
IMPORTANT: Remove any images that are from take-off and landing; also remove any images which are far above or below the average height of the collection.
Run Dense Cloud Reconstruction
Press the 'enter' key to open console
Type 'rotatex-90' to reset the model orientation to nadir
this is important as it will stop the later 'squashing' of images when they are generated in the next step.
Save File and Close V-SfM
Step 2: Open Cloud Compare
Import the Bundle.out file from VSfM
This generates a series of oblique images from the 3-D modelling
If the images appear 'squashed' - meaning they are very long trapezoids and not rectangular trapezoids, go back to V-SfM and try using the 'rotatex' command, changing the angle of rotation, e.g. -90 to 90 degrees.
Step 3: Open Microsoft ICE
Load oblique images from model directory
WARNING: any images which are from different altitudes will alter the scale of the model - ICE does not take into account variation in image footprint size.
Stitch
Save image as .TIF
Step 4: Load image in QGIS
Use QGIS orthophoto projection tool to set control points
Step 5: Open OSGEO shell
Use GDAL2Tile and convert projected .TIF image into Web-ready map format
http://flightriot.com/cmpmvs-and-meshlab-tutorial/
Demo videos